Datasets:

pratyushmaini

/

llm_dataset_inference

like 1

jecture, which asserts that the Weyl tensor should vanish to an order greater than $\frac{n-6}{2}$ at a blow-up point (see [@Schoen3]). The Weyl Vanishing Conjecture has been verified in dimensions $6$ and $7$ by F. Marques [@Marques] and, independently, by Y.Y. Li and L. Zhang [@Li-Zhang1]. Using these results and the positive mass theorem, these authors were able to prove compactness for $n \leq 7$. Moreover, Li and Zhang showed that compactness holds in all dimensions provided that $|W_g(p)| + |\nabla W_g(p)| > 0$ for all $p \in M$. In dimensions $10$ and $11$, it is sufficient to assume that $|W_g(p)| + |\nabla W_g(p)| + |\nabla^2 W_g(p)| > 0$ for all $p \in M$ (see [@Li-Zhang2]). Very recently, M. Khuri, F. Marques and R. Schoen [@Khuri-Marques-Schoen] proved the Weyl Vanishing Conjecture up to dimension $24$. This result, combined with the positive mass theorem, implies the Compactness Conjecture for those dimensions. After proving sharp pointwise estimates, they reduce these questions to showing a certain quadratic form is positive definite. It turns out the quadratic form has negative eigenvalues if $n \geq 25$. In a recent paper [@Brendle1], it was shown that the Compactness Conjecture fails for $n \geq 52$. More precisely, given any integer $n \geq 52$, there exists a smooth Riemannian metric $g$ on $S^n$ such that set of constant scalar curvature metrics in the conformal class of $g$ is non-compact. Moreover, the blowing-up sequences obtained in [@Brendle1] form exactly one bubble. The construction relies on a gluing procedure based on some local model metric. These local models are directions in which the quadratic form of [@Khuri-Marques-Schoen] is negative definite. We refer to [@Brendle2] for a survey of this and related results. In the present paper, we extend these counterexamples to the dimensions $25 \leq n \leq 51$. Our main

jecture, which asserts that the Weyl tensor should vanish to an order greater than $ \frac{n-6}{2}$ at a waste - up item (see [ @Schoen3 ]). The Weyl Vanishing Conjecture has been verified in dimensions $ 6 $ and $ 7 $ by F.   Marques [ @Marques ] and, independently, by Y.Y.   Li and L.   Zhang [ @Li - Zhang1 ]. use these results and the positive batch theorem, these authors were able to prove concentration for $ newton \leq 7$. Moreover, Li and Zhang show that compactness holds in all property provide that $ |W_g(p)| + |\nabla W_g(p)| > 0 $ for all $ p \in M$. In dimensions $ 10 $ and $ 11 $, it is sufficient to assume that $ |W_g(p)| + |\nabla W_g(p)| + |\nabla^2 W_g(p)| > 0 $ for all $ p \in M$ (see [ @Li - Zhang2 ]). Very recently, M.   Khuri, F.   Marques and R.   Schoen [ @Khuri - Marques - Schoen ] prove the Weyl Vanishing Conjecture up to dimension $ 24$. This result, combined with the positivist mass theorem, implies the Compactness Conjecture for those dimensions. After proving crisp pointwise estimates, they reduce these questions to showing a certain quadratic mannequin is positive definite. It turns out the quadratic form has negative eigenvalues if $ n \geq 25$. In a recent paper [ @Brendle1 ], it was shown that the Compactness Conjecture fails for $ n \geq 52$. More precisely, given any integer $ n \geq 52 $, there exists a smooth Riemannian metric $ g$ on $ S^n$ such that stage set of constant scalar curvature metrics in the conformal course of $ g$ is non - compendious. Moreover, the blowing - up sequences obtained in [ @Brendle1 ] form precisely one bubble. The construction relies on a gluing procedure free-base on some local model metric. These local models are directions in which the quadratic form of [ @Khuri - Marques - Schoen ] is negative definite. We consult to [ @Brendle2 ] for a survey of this and related results. In the present paper, we strain these counterexample to the dimension $ 25 \leq n \leq 51$. Our main

jechure, which asserts that uhe Weyl tensor skiuld venish tk an orddr greater than $\frac{n-6}{2}$ at a bpoq-up piint (see [@Schoen3]). The Wehl Vanishpng Conjexturt has been verifixs in dimensiohd $6$ aud $7$ by F. Marques [@Karques] ang, independentlf, cy Y.Y. Li and L. Zhang [@Li-Zhang1]. Using thesq resulys and the posityve kwss fheorem, these authors were able to prove bompactness for $n \leq 7$. Moreover, Li and Zhanh shlwed that compactnfss holds ib alj dimensions orovided that $|W_g(p)| + |\nabma W_g(p)| > 0$ for all $p \in M$. In dimevsionx $10$ and $11$, it iw skxficient to assuie that $|W_g(p)| + |\nabla W_c(p)| + |\nabka^2 W_g(p)| > 0$ for akl $' \in M$ (see [@Li-Zhang2]). Very rerently, M. Khuri, F. Marqugs and R. Scvozn [@Khuri-Marques-Schoen] peoved the Weyu Vavisgiig Donjechurx up to dimsnsion $24$. Thiw result, combined wotr the positive mass ehqorem, implies the Compactness Conjecturt for those dimensions. After proving sharp pointwlse estimwtes, they reduce these questions to showing a cerdain sjadxqtic wirl is positive definite. It turns out the quadrwfiv norm has negativc eigenvalues if $n \gfq 25$. Yn a recent pxper [@Bxsnsle1], it was shown tjat the Compqctness Cjnjevture fails for $n \geq 52$. More precisely, gpven any integer $n \geq 52$, there exirts s smopth Riemannian metric $g$ on $A^n$ such thah set of dunstant scalar cjrvstgre metrics in the conformwl class if $g$ is non-zompsct. Moweover, the blowltg-up sequences obtwined it [@Brendle1] vorm exactly one bubble. The construction religs mn d gluing proccdure based on fome local modgl metric. Thesd local mosels arx directions in which the euadratic focm of [@Khuwi-Maequew-Schoen] ks negative deginite. We refer to [@Vrendle2] for a survcy of fhis and relateb twsults. In the ptesdnt pwpxr, we axtend these couvtefrxampues to uke bioensoons $25 \leq n \leq 51$. Our maih

jecture, which asserts that the Weyl tensor to order greater $\frac{n-6}{2}$ at a Weyl Conjecture has been in dimensions $6$ $7$ by F. Marques [@Marques] and, by Y.Y. Li and L. Zhang [@Li-Zhang1]. Using these results and the positive theorem, these authors were able to prove compactness for $n \leq 7$. Moreover, and showed compactness in all dimensions provided that $|W_g(p)| + |\nabla W_g(p)| > 0$ for all $p \in M$. dimensions $10$ and $11$, it is sufficient to that $|W_g(p)| + |\nabla + |\nabla^2 W_g(p)| > 0$ all \in M$ [@Li-Zhang2]). recently, Khuri, F. Marques R. Schoen [@Khuri-Marques-Schoen] proved the Weyl Vanishing Conjecture up to dimension $24$. This result, combined with the mass theorem, Compactness Conjecture those After sharp pointwise estimates, these questions to showing a certain positive definite. It turns out the quadratic form negative eigenvalues $n \geq 25$. In a recent [@Brendle1], it was shown that the Compactness Conjecture for $n \geq 52$. More precisely, given any integer $n \geq 52$, there exists a metric $g$ on $S^n$ that set of scalar metrics the class of is non-compact. Moreover, the blowing-up sequences obtained in [@Brendle1] form exactly bubble. The construction relies on a gluing procedure based on model These local models directions in which the form [@Khuri-Marques-Schoen] is negative definite. to for this related In the present paper, extend these counterexamples to the $25 \leq n \leq

jecture, which asserts that thE Weyl tensoR shouLd vAniSh To an OrdeR greater than $\frAC{n-6}{2}$ at A blow-up point (see [@Schoen3]). THe WeyL VANishINg conjeCture haS BeEN VerIfIeD in DiMEnSions $6$ And $7$ By F. MarqUes [@Marques] And, InDependently, bY y.Y. li and L. ZhanG [@Li-zhang1]. Using thEse ResultS aNd tHE posiTivE mass TheoreM, These aUthors werE aBLe to prOVe compaCTNeSs foR $n \leq 7$. Moreover, Li anD zhANg showed that coMpactnEsS HoLDS in All Dimensions PrOvideD That $|W_g(p)| + |\NAbLA w_G(p)| > 0$ fOR all $p \in M$. In dimEnsions $10$ and $11$, iT Is sUfficiEnT to ASsume tHat $|W_g(P)| + |\nABla w_g(p)| + |\nabla^2 W_g(p)| > 0$ For aLl $p \in M$ (see [@li-ZhanG2]). very recENtly, M. KhUri, F. MaRquEs aNd R. SCHoEn [@khuRi-mArqUEs-schOEn] pRoved the weYl vanisHing cONJEctuRe uP to dImensIon $24$. This result, ComBineD WitH the pOsitiVe maSs TheorEm, implIes thE COmpactness ConjeCturE for those DimEnSioNs. after PRoving ShaRp pOintwisE estimaTEs, tHeY REDuCe these questions to ShOWInG a certaiN quadrATiC fORm is posiTiVe dEfinITE. It tuRns oUT tHe quadraTic forM HaS nEgative EiGenvalUeS if $N \geQ 25$. In a rECent Paper [@BRendle1], it Was shOWn that the CompaCTness ConjectuRE fAILs FOr $n \gEq 52$. MOre preciselY, givEN any InteGEr $N \geQ 52$, There ExistS a SMoOTh Riemannian metric $g$ On $s^n$ such That sEt of constant sCalar curvaTURE metrics In thE CoNFormal class of $g$ Is non-Compact. MorEOver, the bLowinG-up sequeNces obtaiNED in [@BrendLe1] fOrm ExaCtlY ONe Bubble. The consTRUctiOn Relies oN a gLuing prOceDurE baSed On Some local Model metRiC. THeSe LocAl modELs are dirEcTioNs In wHich tHE quadrAtic fOrm oF [@KHuRI-MaRques-ScHOeN] IS negAtIvE defIniTe. we refEr to [@bRenDle2] for a Survey of tHis ANd reLaTeD resultS. In the present PaPer, we extenD tHesE countERExamples To the dimensions $25 \leq n \leq 51$. OUR main

jecture, which asserts tha t the Weyl tens orsho ul d va nish to an order g r eate r than $\frac{n-6}{2}$ at a b l ow-u p p oint(see [@ S ch o e n3] ). T heWe y lVanis hin g Conje cture hasbee nverified ind im ensions $6 $ a nd $7$ by F.  Ma rques[@ Mar q ues]and , ind epende n tly, b y Y.Y. Li a n d L. Z h ang [@L i - Zh ang1 ]. Using these re s ul t s and the posi tive m as s t h e ore m,these auth or s wer e able t o p r o v e c o mpactness for $n \leq 7$ . Mo reover ,Lia nd Zha ng sh ow e d t hat compact ness holds in all d i mension s provid ed tha t $ |W_ g(p) | +|\n ab l a W _ g( p)| > 0 $ for al l$p \inM$.I n d imen sio ns $ 10$ a nd $11$, it i s s uffi c ien t toassum e th at $|W_ g(p)|+ |\n ab la W_g(p)| + |\ nabl a^2 W_g(p )|>0$fo r all $p \in M$ (s ee [@Li -Zhang2 ] ). V e r y r ecently, M. Khuri, F . Ma rques an d R. S c ho en [@Khuri- Ma rqu es-S c h oen]prov e dthe Weyl Vanis h in gConject ur e up t odim ens ion $ 2 4$.This r esult, c ombin e d with the pos i tive mass the o re m , i m plie s t he Compactn essC onje ctur e f ort hosedimen si o ns . After proving shar ppointw ise e stimates, the y reduce t h e s e questi onst os howing a certa in qu adratic fo r m is pos itive definit e. It tur n s out the qu adr ati c f o r mhas negativee i genv al ues if$n\geq 25 $. In arec en t paper [ @Brendle 1] ,it w asshown that the C omp ac tne ss Co n jectur e fai ls f or $ n \g eq 52$. Mo r e pre ci se ly,giv en anyinte g er$n \geq 52$, the ree xist sasmoothRiemannian me tr ic $g$ on$S ^n$ sucht h at set o f constant scalar curva t ure met ric s intheconformal cl ass of $g $ is no n-comp act.Mo reo v e r, th e bl owi ng -up sequen c e s o btain ed in[@Brend le1] form exactlyo nebubble. The c ons truc t i on re l ie s on a glu i n g procedure bas ed on some l o ca l model me t ric .These l ocal mo delsa re dire ctions in which th equad r a tic form of [ @Khuri-M arques-Sc h oen]i snegat ive defin it e.We re fer to [@B rendl e2] fo ra surv ey of t his andrelated results. In th e pres ent p ape r, we ext end the se counte rexa mples to t hedim ensio ns$ 25 \l eq n \l eq5 1$. O ur m a in

jecture, which_asserts that_the Weyl tensor should_vanish to_an_order greater_than_$\frac{n-6}{2}$ at a_blow-up point (see_[@Schoen3]). The Weyl Vanishing_Conjecture has been_verified_in dimensions $6$ and $7$ by F. Marques [@Marques] and, independently, by Y.Y. Li and L. Zhang_[@Li-Zhang1]._Using these_results_and_the positive mass theorem, these_authors were able to prove_compactness for_$n \leq 7$. Moreover, Li and Zhang showed_that_compactness holds in_all dimensions provided that $|W_g(p)| + |\nabla W_g(p)| >_0$ for all $p \in M$._In dimensions $10$_and_$11$,_it is sufficient to_assume that $|W_g(p)| + |\nabla W_g(p)|_+ |\nabla^2 W_g(p)| > 0$ for_all $p \in M$ (see [@Li-Zhang2]). Very recently,_M. Khuri, F. Marques and R. Schoen [@Khuri-Marques-Schoen] proved_the Weyl Vanishing Conjecture up_to dimension_$24$. This result, combined with_the positive mass_theorem, implies_the Compactness Conjecture_for those dimensions. After proving sharp_pointwise estimates, they_reduce these questions to showing a_certain_quadratic form is_positive_definite._It turns_out the quadratic_form_has negative_eigenvalues_if $n \geq 25$. In a recent_paper_[@Brendle1], it was shown that the Compactness_Conjecture fails for $n_\geq_52$. More precisely, given_any integer $n \geq 52$,_there exists a smooth Riemannian metric_$g$ on_$S^n$ such_that set of constant scalar curvature metrics in the conformal class_of $g$ is non-compact. Moreover, the_blowing-up sequences obtained in_[@Brendle1] form_exactly_one bubble. The_construction_relies on_a gluing procedure based on some local_model metric._These local models are directions in_which the quadratic form_of_[@Khuri-Marques-Schoen] is negative definite. We refer_to [@Brendle2] for a survey of_this and related results. In the_present_paper,_we extend these counterexamples to_the dimensions $25 \leq n \leq_51$. Our main

We also address the question of how to choose the trimming threshold. One extreme possibility is fixed trimming ($b_n=b>0$). Although fixed trimming helps restore asymptotic Gaussianity by forcing the probability weights to be bounded away from zero, this practice is difficult to justify, unless one is willing to re-interpret the estimation and inference result completely. We instead propose to determine the trimming threshold by taking into account the bias and variance of the (trimmed) IPW estimator. We suggest an easy-to-implement method to choose the trimming threshold by minimizing an empirical analogue of the asymptotic mean squared error. This paper relates to a large body of literature on program evaluation and causal inference [@imbens2015causal; @abadie2018econometric; @hernan2018book]. Estimators with inverse weighting are widely used in missing data models [@robins1994estimation; @wooldridge2007inverse] and treatment effect estimation [@hirano2003efficient; @cattaneo2010efficient]. They also feature in settings such as instrumental variables [@abadie2003semiparametric], difference-in-differences [@abadie2005semiparametric], and counterfactual analysis [@dinardo1996labor]. [@khan2010irregular] show that, depending on tail behaviors of the probability weights, the variance bound of the IPW estimator can be infinite, which leads to a slower-than-$\sqrt{n}$ convergence rate. [@sasaki2018ratio] propose a trimming method and a companion sieve-based bias correction technique for conducting inference for moments of ratios, which complement our paper. [@chaudhuri2014heavy] propose a different trimming strategy based on $|DY/e( X)|$ rather than the probability weight, and an inference procedure relying on asymptotic Gaussianity. [-@crump2009dealing] also study the problem of trimming threshold selection, the difference is that their method is based on minimizing a variance term, and hence can lead to a much larger trimming threshold than what we propose. The untrimmed IPW estimator (i.e., $b_n=0$) is a special case of, and the asymptotic distribution is known as the Lévy stable distribution. Stable convergence has been established in many contexts. For example, [@vaynman2014stable] show that stable convergence may arise for the variance targeting estimator, and hence the tail trimming of [@hill2012variance] can be crucial

We also address the question of how to choose the trimming brink. One extreme hypothesis is fixed trimming ($ b_n = b>0 $). Although specify trim helps restore asymptotic Gaussianity by forcing the probability weight to be bounded away from zero, this exercise is unmanageable to justify, unless one is willing to re - understand the estimation and inference result completely. We alternatively propose to determine the trimming threshold by taking into history the bias and variance of the (trimmed) IPW calculator. We suggest an easy - to - enforce method to choose the pare threshold by minimizing an empirical analogue of the asymptotic mean squared error. This paper refer to a large body of literature on program evaluation and causal inference [ @imbens2015causal; @abadie2018econometric; @hernan2018book ]. Estimators with inverse weighting are wide use in missing data models [ @robins1994estimation; @wooldridge2007inverse ] and treatment effect estimation [ @hirano2003efficient; @cattaneo2010efficient ]. They also feature in settings such as implemental variable [ @abadie2003semiparametric ], difference - in - dispute [ @abadie2005semiparametric ], and counterfactual psychoanalysis [ @dinardo1996labor ]. [ @khan2010irregular ] show that, depending on tail behaviors of the probability weights, the variance bound of the IPW estimator can be countless, which leads to a slower - than-$\sqrt{n}$ convergence rate. [ @sasaki2018ratio ] propose a trimming method and a companion sieve - based bias correction technique for conducting inference for moment of ratio, which complement our paper. [ @chaudhuri2014heavy ] propose a different trimming strategy free-base on $ |DY / e (X)|$ rather than the probability weight, and an inference procedure trust on asymptotic Gaussianity. [ -@crump2009dealing ] also learn the problem of trimming doorsill selection, the difference is that their method acting is based on minimizing a variance term, and therefore can lead to a much large trimming threshold than what we nominate. The untrimmed IPW calculator (i.e., $ b_n=0 $) is a special case of, and the asymptotic distribution is known as the Lévy static distribution. Stable convergence has been established in many contexts. For exercise, [ @vaynman2014stable ] prove that stable convergence may get up for the variance targeting calculator, and hence the tail trimming of [ @hill2012variance ] can be crucial

We wlso address the questiok of how to choose the vrimminf threshuld. One extreme possibility ms fuxed ugimming ($b_n=b>0$). Although wixed trilming heops cestore asymptotmd Gausslcnity ny foxcmng the probabikity weighds to be boundad acay from zero, this practice is diffisult to jkstify, unless jne ps wilmpnn to re-interpret the estimation znd inftrence result compketely. We instead propose ho dftermine the trimmlng threshood br taking into account the bias and bariance of the (trimmed) IPW estkmatox. We suggesj zn gasy-to-implemxnt menhod to choosc the tsimming threshold by kinmmizung an empirical analmgue of the asymptjtic mean ssuared error. This paper rglatev to q lxrgt bpdg of lltecature on pdogram evalyation and causal imfqgrnce [@imbens2015czusal; @wbwdie2018econometric; @hernan2018book]. Estimators wpth jnverse weighting are wudely used in missing data modqls [@robins1994estimation; @wooldridge2007inverse] and treatmant ehfdct efgumwtion [@hirano2003efficient; @cattaneo2010efficient]. They wmsp neature in settikgs such as instrukejtsj variables [@acadie2003szjilarametric], differejce-in-disferebces [@abadye2005sekiparametric], and counterfacrual analysif [@dinardo1996labor]. [@khan2010nrregular] shuw tnat, drpending on tail behaviurs kf the probwbility wskghts, the varianze noutd of tht IPW estimator cag be infiiite, chich lexds jo a sljwer-than-$\sqgt{n}$ convergence rate. [@saswki2018rajio] prmpose a trlmming method and a companion smxve-based bias cmrrvction teehniquc for conductind inference fot moments of rxtios, whicg complxment our pa[er. [@chaudhuri2014vgavy] propose e differegt teimmung strxgegy based on $|CY/e( X)|$ ratktr than thw probability weigmt, ana an inference pxuxedure relying pn xsyipnotmc Gatvsianity. [-@crukp2009dexlivb] alsu study the orobkem of trimming threvhols selection, the digfcrence is that thqir method is based on minimizijg a tarianre terk, agd hence can lead to a much ladger trimlinn threshold tran qhat we proppse. The untrimmed IPW estimator (i.e., $b_n=0$) is a special case of, and the asymptotic cistributioi is kgown as tve Lévy stable distrivution. Stable conyergence has been estaglisheg in lany contexts. For example, [@vaynman2014stable] show that stable convergence may qrise how the variahce yargeniny extimatjr, aid hence the tail nrimming of [@hill2012variance] can be rrucial

We also address the question of how the threshold. One possibility is fixed helps asymptotic Gaussianity by the probability weights be bounded away from zero, this is difficult to justify, unless one is willing to re-interpret the estimation and result completely. We instead propose to determine the trimming threshold by taking into the and of (trimmed) IPW estimator. We suggest an easy-to-implement method to choose the trimming threshold by minimizing an analogue of the asymptotic mean squared error. This relates to a large of literature on program evaluation causal [@imbens2015causal; @abadie2018econometric; Estimators inverse are widely used missing data models [@robins1994estimation; @wooldridge2007inverse] and treatment effect estimation [@hirano2003efficient; @cattaneo2010efficient]. They also feature in settings such instrumental variables [@abadie2005semiparametric], and analysis [@khan2010irregular] that, depending on of the probability weights, the variance IPW estimator can be infinite, which leads to slower-than-$\sqrt{n}$ convergence [@sasaki2018ratio] propose a trimming method and companion sieve-based bias correction technique for conducting inference moments of ratios, which complement our paper. [@chaudhuri2014heavy] propose a different trimming strategy based on rather than the probability and an inference relying asymptotic [-@crump2009dealing] study the of trimming threshold selection, the difference is that their method is on minimizing a variance term, and hence can lead to larger threshold than what propose. The untrimmed IPW (i.e., is a special case the distribution the stable Stable convergence has been in many contexts. For example, show that stable convergence targeting estimator, and hence the tail trimming of can be crucial

We also address the question oF how to chooSe the TriMmiNg ThreSholD. One extreme posSIbilIty is fixed trimming ($b_n=b>0$). ALthouGh FIxed TRiMming Helps reSToRE AsyMpToTic gaUSsIanitY by Forcing The probabiLitY wEights to be boUNdEd away from ZerO, this practicE is DifficUlT to JUstifY, unLess oNe is wiLLing to Re-interprEt THe estiMAtion anD INfErenCe result completelY. we INstead propose tO deterMiNE tHE TriMmiNg thresholD bY takiNG into acCOuNT THe bIAs and variance Of the (trimmeD) iPW EstimaToR. We SUggest An easY-tO-ImpLement methoD to cHoose the tRimminG ThreshoLD by miniMizing An eMpiRicaL AnAlOguE oF The ASyMptOTic Mean squaReD eRror. THis pAPER RelaTes To a lArge bOdy of literatuRe oN proGRam EvaluAtion And cAuSal inFerencE [@imbeNs2015Causal; @abadie2018ecoNomeTric; @hernaN2018boOk]. estImAtors WIth invErsE weIghting Are wideLY usEd IN MIsSing data models [@robiNs1994ESTiMation; @woOldridGE2007iNvERse] and trEaTmeNt efFECt estImatIOn [@Hirano2003efFicienT; @CaTtAneo2010effIcIent]. ThEy AlsO feAture IN setTings sUch as insTrumeNTal variables [@abADie2003semiparameTRiC], DIfFErenCe-iN-differenceS [@abaDIe2005seMipaRAmEtrIC], and cOunteRfACtUAl analysis [@dinardo1996laBoR]. [@khan2010iRreguLar] show that, dePending on tAIL BehaviorS of tHE pRObability weighTs, the Variance boUNd of the IpW estImator caN be infiniTE, Which leaDs tO a sLowEr-tHAN-$\sQrt{n}$ convergenCE Rate. [@SaSaki2018ratIo] pRopose a TriMmiNg mEthOd And a compaNion sievE-bAsEd BiAs cOrrecTIon technIqUe fOr ConDuctiNG inferEnce fOr moMeNtS Of rAtios, whICh COMpleMeNt Our pApeR. [@cHaudhUri2014hEAvy] Propose A differenT trIMminG sTrAtegy baSed on $|DY/e( X)|$ rathEr Than the proBaBilIty weiGHT, and an inFerence procedure relying ON asymptOtiC GausSianIty. [-@crump2009dEalIng] alsO stUDy the pRoblem Of triMmIng THReshoLD SeLecTiOn, the diffeRENce Is thaT tHeir Method iS based on minimizing A VarIance term, and hEncE can LEAd To a MUcH LarGeR TriMMIng threshold thaN what we proPoSE. THe untrimmeD iPW EsTimator (I.e., $b_n=0$) is a SpeciAL case of, And the asyMptotic diStRibuTIOn iS known as thE Lévy staBle distriBUtion. sTaBle coNveRgence HaS beEn estAblishED in Many cOntextS. FOr examPle, [@vaYnMan2014stablE] show that stable convergeNce may Arise For The varianCe tARgeTing estimAtor, And hence thE taIl tRimmiNg oF [@Hill2012vAriaNCe] Can BE crucIal

We also address the questi on of howto ch oos e t he tri mmin g threshold. O n e ex treme possibility is f ixedtr i mmin g ( $b_n= b>0$).A lt h o ugh f ix edtr i mm ing h elp s resto re asympto tic G aussianity b y f orcing the pr obability we igh ts tobe bo u ndedawa y fro m zero , thispracticeis diffic u lt to j u s ti fy,unless one is wil l in g to re-interpr et the e s ti m a tio n a nd inferen ce resu l t compl e te l y . We instead propo se to deter m ine the t ri mmi n g thre shold b y ta king into a ccou nt the bi as and varianc e of the (trim med ) I PW e s ti ma tor .W e s u gg est aneasy-to- im pl ement met h o d to c hoo se t he tr imming thresh old bym ini mizin g anempi ri cal a nalogu e ofth e asymptotic me an s quared er ror . Th is pape r relat estoa large body o f li te r a t ur e on program evalu at i o nand caus al inf e re nc e [@imben s2 015 caus a l ; @ab adie 2 01 8econome tric;@ he rn an2018b oo k]. Es ti mat ors with inve rse we ightingare w i dely used in m i ssing data mo d el s [@ r obin s19 94estimatio n; @ w oold ridg e 20 07i n verse ] and t r ea t ment effect estimat io n [@hi rano2 003efficient; @cattaneo 2 0 1 0efficie nt]. Th e y also feature in s ettings su c h as ins trume ntal var iables [@ a b adie2003 sem ipa ram etr i c ], difference-i n - diff er ences [ @ab adie200 5se mip ara met ri c], and c ounterfa ct ua lan aly sis [ @ dinardo1 99 6la bo r]. [@kh a n2010i rregu lar] s ho w th at, dep e nd i n g on t ai l be hav io rs of the pro babilit y weights , t h e va ri an ce boun d of the IPWes timator ca nbeinfini t e , whichleads to a slower-than- $ \sqrt{n }$conve rgen ce rate.[@s asaki2 018 r atio]propos e a t ri mmi n g meth o d a ndacompanions i eve -base dbias correc tion technique for con ducting infer enc e fo r mo men t so f r at i os, w hich complement our paper .[ @c haudhuri20 1 4he av y] prop ose a d iffer e nt trim ming stra tegy base don $ | D Y/e ( X)|$ rat her than the prob a bilit y w eight , a nd anin fer enceproced u rerelyi ng onas ymptot ic Ga us sianity. [-@crump2009dealing] a lso st udy t heproblem o f t r imm ing thres hold selection , t hediffe ren c e isthat th eir metho d is based onm in imi z i ng a variance t e rm, andhen c e canlead to a much larger trimming thres hold t han wh a t we p ropose. The u ntr im m e d IPW es ti mator (i.e. , $b_n=0 $) is aspecia l case of, an d th e asymp toti c d istributi onis known a sth e Lévystab le distr ibutio n . St a b le convergence h as be e n esta b lis hed i nmany co n text s. For exa mple, [@vay nman20 14st able] show t ha t stab leco nvergencem ay arisefor t he vari an ce t arg etingesti m a tor,andhe nce the tail t ri m mi ng of[@hi ll201 2v aria nce] canb e crucia l

We also_address the_question of how to_choose the_trimming_threshold. One_extreme_possibility is fixed_trimming ($b_n=b>0$). Although_fixed trimming helps restore_asymptotic Gaussianity by_forcing_the probability weights to be bounded away from zero, this practice is difficult to_justify,_unless one_is_willing_to re-interpret the estimation and_inference result completely. We instead_propose to_determine the trimming threshold by taking into account_the_bias and variance_of the (trimmed) IPW estimator. We suggest an easy-to-implement_method to choose the trimming threshold_by minimizing an_empirical_analogue_of the asymptotic mean_squared error. This paper relates to a_large body of literature on program_evaluation and causal inference [@imbens2015causal; @abadie2018econometric; @hernan2018book]._Estimators with inverse weighting are widely_used in missing data models_[@robins1994estimation; @wooldridge2007inverse]_and treatment effect estimation [@hirano2003efficient;_@cattaneo2010efficient]. They also_feature in_settings such as_instrumental variables [@abadie2003semiparametric], difference-in-differences [@abadie2005semiparametric], and_counterfactual analysis [@dinardo1996labor]._[@khan2010irregular] show that, depending on tail_behaviors_of the probability_weights,_the_variance bound_of the IPW_estimator_can be_infinite,_which leads to a slower-than-$\sqrt{n}$ convergence_rate._[@sasaki2018ratio] propose a trimming method and a_companion sieve-based bias correction_technique_for conducting inference for_moments of ratios, which complement_our paper. [@chaudhuri2014heavy] propose a different_trimming strategy_based on_$|DY/e( X)|$ rather than the probability weight, and an inference procedure_relying on asymptotic Gaussianity. [-@crump2009dealing] also_study the problem of_trimming threshold_selection,_the difference is_that_their method_is based on minimizing a variance term,_and hence_can lead to a much larger_trimming threshold than what_we_propose. The untrimmed IPW estimator (i.e., $b_n=0$)_is a special case of, and_the asymptotic distribution is known_as_the_Lévy stable distribution. Stable convergence_has been established in many contexts._For example, [@vaynman2014stable]_show that stable convergence may arise for_the_variance targeting estimator, and hence the_tail_trimming of [@hill2012variance] can be crucial

alpha_d = \exp{2 \pi / d}$ is the complex $d$th root of unity. d=4 === The bases of this Appendix present explicitly the result of construction described in section 3.2 of the main text. $$\label{dim4a} B_4 = \left(\begin{array}{cccc} 1&0&0&0\\ 0&1&0&0\\ 0&0&1&0\\ 0&0&0&1 \end{array} \right) = \left\{\begin{array}{c} |0\rangle\\ |1 \rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0\rangle\\ |1\rangle \end{array} \right\},$$ $$B_0 = \frac{1}{2} \left(\begin{array}{cccc} 1&1&1&1\\ 1&-1&1&-1\\ 1&1&-1&-1\\ 1&-1&-1&1 \end{array} \right) = \left\{\begin{array}{c} |0_0\rangle\\ |1_0\rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0_0\rangle\\ |1_0\rangle \end{array} \right\},$$ $$B_1 = \frac{1}{2} \left( \begin{array}{cccc} 1&1&1&1\\ i&-i&i&-i\\ i&i&-i&-i\\ -1&1&1&-1 \end{array}\right) = \left\{\begin{array}{c} |0_1\rangle\\ |1_1\rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0_1\rangle\\ |1_1\rangle \end{array} \right\},$$ $$\begin{aligned} B_2 & = & \frac{1}{2}

alpha_d = \exp{2 \pi / d}$ is the complex $ d$th root of unity. d=4 = = = The bases of this Appendix present explicitly the result of structure trace in section 3.2 of the main textbook. $ $ \label{dim4a } B_4 = \left(\begin{array}{cccc } 1&0&0&0\\ 0&1&0&0\\ 0&0&1&0\\ 0&0&0&1 \end{array } \right) = \left\{\begin{array}{c } |0\rangle\\ |1 \rangle \end{array } \right\ } \otimes \left\{\begin{array}{c } |0\rangle\\ |1\rangle \end{array } \right\},$$ $ $ B_0 = \frac{1}{2 } \left(\begin{array}{cccc } 1&1&1&1\\ 1&-1&1&-1\\ 1&1&-1&-1\\ 1&-1&-1&1 \end{array } \right) = \left\{\begin{array}{c } |0_0\rangle\\ |1_0\rangle \end{array } \right\ } \otimes \left\{\begin{array}{c } |0_0\rangle\\ |1_0\rangle \end{array } \right\},$$ $ $ B_1 = \frac{1}{2 } \left ( \begin{array}{cccc } 1&1&1&1\\ i&-i&i&-i\\ i&i&-i&-i\\ -1&1&1&-1 \end{array}\right) = \left\{\begin{array}{c } |0_1\rangle\\ |1_1\rangle \end{array } \right\ } \otimes \left\{\begin{array}{c } |0_1\rangle\\ |1_1\rangle \end{array } \right\},$$ $ $ \begin{aligned } B_2 & = & \frac{1}{2 }

alpja_d = \exp{2 \pi / d}$ is the cooplex $d$th root of unitb. d=4 === The bzses of ghis Appendix present explicmtly the eesult of construction describef in secrion 3.2 of the mamh text. $$\label{dij4w} B_4 = \leyt(\uegin{array}{cccc} 1&0&0&0\\ 0&1&0&0\\ 0&0&1&0\\ 0&0&0&1 \ekd{array} \righd) = \left\{\begin{arraf}{c} |0\faugle\\ |1 \rangle \end{array} \right\} \otimes \left\{\bedin{arrau}{c} |0\gangle\\ |1\rangle \enq{arrsr} \riggn\},$$ $$B_0 = \nrac{1}{2} \left(\begin{array}{cccc} 1&1&1&1\\ 1&-1&1&-1\\ 1&1&-1&-1\\ 1&-1&-1&1 \end{array} \rifht) = \left\{\uegin{array}{c} |0_0\rangke\\ |1_0\rangle \end{array} \right\} \otimed \lefh\{\begin{array}{c} |0_0\rangle\\ |1_0\gangle \end{artzy} \rytht\},$$ $$B_1 = \frac{1}{2} \left( \cegin{array}{becc} 1&1&1&1\\ i&-i&i&-i\\ i&i&-i&-i\\ -1&1&1&-1 \gnd{array}\right) = \left\{\begin{array}{c} |0_1\rangue\\ |1_1\ranyle \end{array} \tntht\} \ljimes \left\{\begmn{arraj}{c} |0_1\rangle\\ |1_1\ranglc \vnd{arraf} \right\},$$ $$\brgin{aligned} B_2 & = & \nrac{1}{2}

alpha_d = \exp{2 \pi / d}$ is $d$th of unity. === The bases the of construction described section 3.2 of main text. $$\label{dim4a} B_4 = \left(\begin{array}{cccc} 0&1&0&0\\ 0&0&1&0\\ 0&0&0&1 \end{array} \right) = \left\{\begin{array}{c} |0\rangle\\ |1 \rangle \end{array} \right\} \otimes |0\rangle\\ |1\rangle \end{array} \right\},$$ $$B_0 = \frac{1}{2} \left(\begin{array}{cccc} 1&1&1&1\\ 1&-1&1&-1\\ 1&1&-1&-1\\ 1&-1&-1&1 \end{array} = |0_0\rangle\\ \end{array} \otimes \left\{\begin{array}{c} |0_0\rangle\\ |1_0\rangle \end{array} \right\},$$ $$B_1 = \frac{1}{2} \left( \begin{array}{cccc} 1&1&1&1\\ i&-i&i&-i\\ i&i&-i&-i\\ -1&1&1&-1 \end{array}\right) \left\{\begin{array}{c} |0_1\rangle\\ |1_1\rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0_1\rangle\\ \end{array} \right\},$$ $$\begin{aligned} B_2 = & \frac{1}{2}

alpha_d = \exp{2 \pi / d}$ is the complex $d$Th root of unIty. d=4 === THe bAseS oF thiS AppEndix present exPLiciTly the result of construcTion dEsCRibeD In SectiOn 3.2 of the MAiN TExt. $$\LaBeL{diM4a} b_4 = \LeFt(\begIn{aRray}{cccC} 1&0&0&0\\ 0&1&0&0\\ 0&0&1&0\\ 0&0&0&1 \end{array} \rIghT) = \lEft\{\begin{arraY}{C} |0\rAngle\\ |1 \ranglE \enD{array} \right\} \oTimEs \left\{\BeGin{ARray}{c} |0\RanGle\\ |1\raNgle \enD{Array} \rIght\},$$ $$B_0 = \frac{1}{2} \LeFT(\begin{ARray}{cccC} 1&1&1&1\\ 1&-1&1&-1\\ 1&1&-1&-1\\ 1&-1&-1&1 \ENd{ArraY} \right) = \left\{\begin{arRAy}{C} |0_0\Rangle\\ |1_0\rangle \enD{array} \RiGHt\} \OTImeS \leFt\{\begin{arrAy}{C} |0_0\rangLE\\ |1_0\rangle \ENd{ARRAy} \rIGht\},$$ $$B_1 = \frac{1}{2} \left( \bEgin{array}{ccCC} 1&1&1&1\\ i&-i&I&-i\\ i&i&-i&-i\\ -1&1&1&-1 \EnD{arRAy}\righT) = \left\{\BeGIn{aRray}{c} |0_1\rangle\\ |1_1\RangLe \end{arraY} \right\} \OTimes \leFT\{\begin{aRray}{c} |0_1\rAngLe\\ |1_1\rAnglE \EnD{aRraY} \rIGht\},$$ $$\BEgIn{aLIgnEd} B_2 & = & \frac{1}{2}

alpha_d = \exp{2 \pi / d}$ is the co mplex $d $th r ootof u nity. d=4 === The bases of this Appendi x pre se n t ex p li citly the re s ul t ofco ns tru ct i on desc rib ed in s ection 3.2 of t he main text . $$\label{d im4 a} B_4 = \le ft( \begin {a rra y }{ccc c}1&0&0 &0\\ 0 & 1&0&0\ \ 0&0&1&0 \\ 0&0&0& 1 \end{a r r ay } \r ight) = \left\{\b e gi n {array}{c} |0\ rangle \\ |1 \ ran gle \end{arra y} \rig h t\} \ot i me s \ lef t \{\begin{arra y}{c} |0\ra n gle \\ |1\ ra ngl e \end{ array }\ rig ht\},$$ $$ B_0= \frac{1 }{2} \ l eft(\be g in{arra y}{ccc c}1&1 &1&1 \ \1& -1& 1& - 1\\ 1& 1&- 1 &-1 \\ 1&-1& -1 &1 \end {arr a y } \rig ht) = \ left\ {\begin{array }{c } |0 _ 0\r angle \\ |1 _0\r an gle \ end{ar ray}\r ight\} \otimes\lef t\{\begin {ar ra y}{ c} |0_0 \ rangle \\|1_ 0\rangl e \end{ a rra y} \ r ig ht\},$$ $$B_1 = \ fr a c {1 }{2} \le ft( \b e gi n{ a rray}{cc cc } 1 &1&1 & 1 \\ i& -i&i & -i \\ i&i&- i&-i\\ -1 &1 &1&-1 \ en d{arra y} \ri ght ) = \ l eft\ {\begi n{array} {c} | 0 _1\rangle\\ |1 _ 1\rangle \end { ar r a y} \rig ht\ } \otimes \ left \ {\be gin{ a rr ay} { c} |0 _1\ra ng l e\ \ |1_1\rangle \end{a rr ay} \r ight\ },$$ $$\begi n{aligned} B _ 2 & = &\fra c {1 } {2}

alpha_d =_\exp{2 \pi_/ d}$ is the_complex $d$th_root_of unity. d=4 === The_bases_of this Appendix_present explicitly the_result of construction described_in section 3.2_of_the main text. $$\label{dim4a} B_4 = \left(\begin{array}{cccc} 1&0&0&0\\ 0&1&0&0\\ 0&0&1&0\\ 0&0&0&1 \end{array} \right) = \left\{\begin{array}{c} |0\rangle\\ |1 \rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0\rangle\\ |1\rangle \end{array} \right\},$$ $$B_0 = \frac{1}{2} \left(\begin{array}{cccc} 1&1&1&1\\ 1&-1&1&-1\\ 1&1&-1&-1\\ 1&-1&-1&1 \end{array} \right) = \left\{\begin{array}{c} |0_0\rangle\\ |1_0\rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0_0\rangle\\ |1_0\rangle \end{array} \right\},$$ $$B_1 = \frac{1}{2} \left( \begin{array}{cccc} 1&1&1&1\\ i&-i&i&-i\\ i&i&-i&-i\\ -1&1&1&-1 \end{array}\right) = \left\{\begin{array}{c} |0_1\rangle\\ |1_1\rangle \end{array} \right\} \otimes \left\{\begin{array}{c} |0_1\rangle\\ |1_1\rangle \end{array} \right\},$$ $$\begin{aligned} B_2 & = & \frac{1}{2}

}\end{aligned}$$ since both square brackets in the last equality in are bounded operators and by hypothesis at least one of them is compact. Employing again completes the proof of. Equation extends [@RS78 Problem 73(a), p. 373], since $B$ is not assumed to be symmetric. In a completely analogous manner one proves membership of the operators in in the Schatten–von Neumann classes ${{\mathcal B}}_p({{\mathcal H}})$, $p{\geqslant}1$; we omit further details. [r3.6]{} We conclude this section by recalling a well-known example, where $A$ and $B$ are self-adjoint, $B$ is $A$-form bounded and even $A$-form compact, but $B$ is not $A$-bounded (let alone $A$-compact): Denote by ${{\mathcal R}}$ the class of Rollnik potentials in ${{\mathbb{R}}}^3$, that is, $${{\mathcal R}}= \bigg\{V\colon{{\mathbb{R}}}^3 \to {{\mathbb{C}}}\,\bigg| \int_{{{\mathbb{R}}}^6} d^3x \, d^3y \, {\frac}{|V(x)| |V(y)|}{|x-y|^2} < \infty\bigg\},$$ and by $H_0$ the $L^2({{\mathbb{R}}};d^3x)$-realization of (minus) the Laplacian $-\Delta$ defined on the Sobolev space $H^{2,2}({{\mathbb{R}}}^3)$. Then there exist potentials $0 \leq V_0 \in L^1({{\mathbb{R}}}^3;d^3x)\cap{{\mathcal R}}$ such that $$V_0^{1/2} \big(H_0 + I_{L^2({{\mathbb{R}}}^3;d^3x)}\big)^{-1/2} \in {{\mathcal B}}_4\big(L^2({{\mathbb{R}}}^3;d^3x)\big),$$ (cf. Simon [@Si71], Theorem I.22 and Example 4 in Sect. I.6) and hence $V_0$ is $H_0$-form compact, but $${\text{\rm{dom}}}(V_0) \cap {\text{\rm{dom}}}(H_0) =\{0\},$$ and thus $V_0$ is not $H_0

} \end{aligned}$$ since both square brackets in the last equality in are bounded operator and by guess at least one of them is compact. Employing again dispatch the validation of. Equation extends [ @RS78 Problem   73(a), p.   373 ], since $ B$ is not assumed to be symmetric. In a wholly analogous manner one rise membership of the operator in in the Schatten – von Neumann classes $ { { \mathcal B}}_p({{\mathcal H}})$, $ p{\geqslant}1 $; we omit further detail. [ r3.6 ] { } We conclude this section by recall a well - know example, where $ A$ and $ B$ are self - adjoint, $ B$ is $ A$ -form bounded and even $ A$ -form compact, but $ B$ is not $ A$ -bounded (let alone $ A$ -compact ): announce by $ { { \mathcal R}}$ the class of Rollnik potentials in $ { { \mathbb{R}}}^3 $, that is, $ $ { { \mathcal R}}= \bigg\{V\colon{{\mathbb{R}}}^3 \to { { \mathbb{C}}}\,\bigg| \int_{{{\mathbb{R}}}^6 } d^3x \, d^3y \, { \frac}{|V(x)| |V(y)|}{|x - y|^2 } < \infty\bigg\},$$ and by $ H_0 $ the $ L^2({{\mathbb{R}}};d^3x)$-realization of (minus) the Laplacian $ -\Delta$ define on the Sobolev space $ H^{2,2}({{\mathbb{R}}}^3)$. Then there exist potentials $ 0 \leq V_0 \in L^1({{\mathbb{R}}}^3;d^3x)\cap{{\mathcal R}}$ such that $ $ V_0^{1/2 } \big(H_0 + I_{L^2({{\mathbb{R}}}^3;d^3x)}\big)^{-1/2 } \in { { \mathcal B}}_4\big(L^2({{\mathbb{R}}}^3;d^3x)\big),$$ (californium.   Simon [ @Si71 ], Theorem I.22 and Example 4 in Sect.   I.6) and hence $ V_0 $ is $ H_0$-form compendious, but $ $ { \text{\rm{dom}}}(V_0) \cap { \text{\rm{dom}}}(H_0) = \{0\},$$ and thus $ V_0 $ is not $ H_0

}\end{wligned}$$ since both squart brackets in the last exualitg in are bounded operators and by hy'othwsis qt least one of them ir compact. Employibg ajain completes tis proof of. Equzbion zxvends [@RS78 Problek 73(a), p. 373], since $B$ is not assukea co be symmetric. In a completely analodous mamnfr one proves iembtrsrip kf the operators in in the Schatteh–von Nelmann classes ${{\matncal B}}_p({{\mathcal H}})$, $p{\geqslant}1$; we lmit further detaips. [r3.6]{} We concoude rhis section by recallpug a well-knkwn example, where $A$ and $B$ are sdlf-adloint, $B$ is $Q$-firm tounded and even $A$-form compagn, but $B$ is not $A$-bounded (let sloie $A$-xompact): Denote by ${{\matical R}}$ the class of Tollnik podeutials in ${{\mathbb{R}}}^3$, thar us, $${{\majhcal R}}= \bktg\{V\zolkn{{\kafhbb{R}}}^3 \ho {{\jathbb{C}}}\,\bigf| \int_{{{\mathbb{R}}}^6} d^3x \, d^3y \, {\frac}{|V(x)| |V(y)|}{|x-u|^2} < \pmfty\bigg\},$$ and by $H_0$ ehq $L^2({{\mathbb{R}}};d^3x)$-realization of (minus) the Laklacizn $-\Delta$ defined on the Sobolev space $H^{2,2}({{\mathbh{R}}}^3)$. Then trere exist potentials $0 \leq V_0 \in L^1({{\mathbb{R}}}^3;d^3x)\cap{{\mathwal R}}$ ruck that $$C_0^{1/2} \hig(H_0 + I_{L^2({{\mathbb{R}}}^3;d^3x)}\big)^{-1/2} \in {{\mathcal B}}_4\big(L^2({{\mathbb{R}}}^3;d^3v)\gib),$$ (bf. Simon [@Si71], Theorei I.22 and Exakppe 4 in Sect. I.6) and hence $V_0$ js $H_0$-form compact, bkt $${\text{\tm{dom}}}(V_0) \cap {\text{\wm{dok}}}(H_0) =\{0\},$$ and thus $V_0$ is not $H_0

}\end{aligned}$$ since both square brackets in the in bounded operators by hypothesis at compact. again completes the of. Equation extends Problem 73(a), p. 373], since $B$ not assumed to be symmetric. In a completely analogous manner one proves membership the operators in in the Schatten–von Neumann classes ${{\mathcal B}}_p({{\mathcal H}})$, $p{\geqslant}1$; we further [r3.6]{} conclude section by recalling a well-known example, where $A$ and $B$ are self-adjoint, $B$ is $A$-form bounded even $A$-form compact, but $B$ is not $A$-bounded alone $A$-compact): Denote by R}}$ the class of Rollnik in that is, R}}= \to \int_{{{\mathbb{R}}}^6} d^3x \, \, {\frac}{|V(x)| |V(y)|}{|x-y|^2} < \infty\bigg\},$$ and by $H_0$ the $L^2({{\mathbb{R}}};d^3x)$-realization of (minus) the Laplacian $-\Delta$ defined on Sobolev space there exist $0 V_0 L^1({{\mathbb{R}}}^3;d^3x)\cap{{\mathcal R}}$ such \big(H_0 + I_{L^2({{\mathbb{R}}}^3;d^3x)}\big)^{-1/2} \in {{\mathcal B}}_4\big(L^2({{\mathbb{R}}}^3;d^3x)\big),$$ Theorem I.22 and Example 4 in Sect. I.6) hence $V_0$ $H_0$-form compact, but $${\text{\rm{dom}}}(V_0) \cap {\text{\rm{dom}}}(H_0) and thus $V_0$ is not $H_0

}\end{aligned}$$ since both square Brackets in The laSt eQuaLiTy in Are bOunded operatorS And bY hypothesis at least one oF them Is COmpaCT. EMployIng agaiN CoMPLetEs ThE prOoF Of. equatIon Extends [@rS78 Problem 73(a), P. 373], siNcE $B$ is not assumED tO be symmetrIc. IN a completely AnaLogous MaNneR One prOveS membErship OF the opErators in In THe SchaTTen–von NEUMaNn clAsses ${{\mathcal B}}_p({{\matHCaL h}})$, $p{\geqslant}1$; we omIt furtHeR DeTAIls. [R3.6]{} We Conclude thIs SectiON by recaLLiNG A WelL-Known example, wHere $A$ and $B$ arE SelF-adjoiNt, $b$ is $a$-Form boUnded AnD EveN $A$-form compaCt, buT $B$ is not $A$-bOunded (LEt alone $a$-Compact): denote By ${{\mAthCal R}}$ THe ClAss Of rOllNIk PotENtiAls in ${{\matHbB{R}}}^3$, That iS, $${{\matHCAL r}}= \bigG\{V\cOlon{{\MathbB{R}}}^3 \to {{\mathbb{C}}}\,\biGg| \iNt_{{{\maTHbb{r}}}^6} d^3x \, d^3y \, {\Frac}{|V(X)| |V(y)|}{|x-Y|^2} < \iNfty\bIgg\},$$ and By $H_0$ thE $L^2({{\Mathbb{R}}};d^3x)$-realizAtioN of (minus) tHe LApLacIaN $-\DeltA$ DefineD on The sobolev Space $H^{2,2}({{\mAThbB{R}}}^3)$. tHEN tHere exist potentialS $0 \lEQ v_0 \iN L^1({{\mathbb{r}}}^3;d^3x)\cap{{\MAtHcAL R}}$ such thAt $$v_0^{1/2} \biG(H_0 + I_{L^2({{\MAThbb{R}}}^3;D^3x)}\biG)^{-1/2} \In {{\Mathcal B}}_4\Big(L^2({{\maTHbB{R}}}^3;D^3x)\big),$$ (cf. siMon [@Si71], THeOreM I.22 aNd ExaMPle 4 iN Sect. I.6) And hence $v_0$ is $H_0$-fORm compact, but $${\teXT{\rm{dom}}}(V_0) \cap {\texT{\Rm{DOM}}}(H_0) =\{0\},$$ ANd thUs $V_0$ Is not $H_0

}\end{aligned}$$ since bot h square b racke tsinth e la st e quality in are boun ded operators and by h ypoth es i s at le ast o ne of t h em i s c om pa ct. E m pl oying ag ain com pletes the pr oo f of. Equat i on extends [ @RS 78 Problem 7 3(a ), p.37 3], since $B $ isnot as s umed t o be symm et r ic. I n a comp l e te ly a nalogous manner o n ep roves membersh ip ofth e o p e rat ors in in the S chatt e n–von N e um a n n cl a sses ${{\math cal B}}_p({ { \ma thcalH} })$ , $p{\g eqsla nt } 1$; we omit fu rthe r details . [r3 . 6]{} We conclud e this se cti on b y r ec all in g aw el l-k n own example ,wh ere $ A$ a n d $ B$ a reself -adjo int, $B$ is $ A$- form bou ndedand e ven$A $-for m comp act,bu t $B$ is not $A $-bo unded (le t a lo ne$A $-com p act):Den ote by ${{ \mathca l R} }$ t h eclass of Rollnik p ot e n ti als in $ {{\mat h bb {R } }}^3$, t ha t i s, $ $ { {\mat hcal R} }= \bigg \{V\co l on {{ \mathbb {R }}}^3\t o { {\m athbb { C}}} \,\big g| \int_ {{{\m a thbb{R}}}^6} d ^ 3x \, d^3y \, {\ f r ac } {|V( x)| |V(y)|}{|x -y|^ 2 } <\inf t y\ big g \},$$ andby $H _ 0$ the $L^2({{\math bb {R}}}; d^3x) $-realization of (minus ) t he Lapla cian $- \ Delta$ defined on t he Sobolev space $H ^{2,2 }({{\mat hbb{R}}}^ 3 ) $. Thenthe reexi stp o te ntials $0 \le q V_0\i n L^1({ {\m athbb{R }}} ^3; d^3 x)\ ca p{{\mathc al R}}$su ch t ha t $ $V_0^ { 1/2} \bi g( H_0 + I_ {L^2( { {\math bb{R} }}^3 ;d ^3 x )}\ big)^{- 1 /2 } \in { {\ math cal B }}_4\ big( L ^2( {{\math bb{R}}}^3 ;d^ 3 x)\b ig ), $$ (cf.  Simon [@Si71 ], Theorem I .2 2 a nd Exa m p le 4 inSect. I.6) and hence $V _ 0$ is $ H_0 $-for m co mpact, bu t $ ${\tex t{\ r m{dom} }}(V_0 ) \ca p{\t e x t{\rm { d om }}} (H _0) =\{0\} , $ $ a nd th us $V_ 0$ is n ot $H_0

}\end{aligned}$$ since_both square_brackets in the last_equality in_are_bounded operators_and_by hypothesis at_least one of_them is compact. Employing_again completes the_proof_of. Equation extends [@RS78 Problem 73(a), p. 373], since $B$ is not assumed to be symmetric. In a_completely_analogous manner_one_proves_membership of the operators in_in the Schatten–von Neumann classes_${{\mathcal B}}_p({{\mathcal_H}})$, $p{\geqslant}1$; we omit further details. [r3.6]{} We conclude_this_section by recalling_a well-known example, where $A$ and $B$ are self-adjoint,_$B$ is $A$-form bounded and even_$A$-form compact, but_$B$_is_not $A$-bounded (let alone_$A$-compact): Denote by ${{\mathcal R}}$ the_class of Rollnik potentials in ${{\mathbb{R}}}^3$,_that is, $${{\mathcal R}}= \bigg\{V\colon{{\mathbb{R}}}^3 \to {{\mathbb{C}}}\,\bigg| \int_{{{\mathbb{R}}}^6}_d^3x \, d^3y \, {\frac}{|V(x)| |V(y)|}{|x-y|^2}_< \infty\bigg\},$$ and by $H_0$_the $L^2({{\mathbb{R}}};d^3x)$-realization_of (minus) the Laplacian $-\Delta$_defined on the_Sobolev space_$H^{2,2}({{\mathbb{R}}}^3)$. Then there_exist potentials $0 \leq V_0 \in_L^1({{\mathbb{R}}}^3;d^3x)\cap{{\mathcal R}}$ such_that $$V_0^{1/2} \big(H_0 + I_{L^2({{\mathbb{R}}}^3;d^3x)}\big)^{-1/2} \in_{{\mathcal_B}}_4\big(L^2({{\mathbb{R}}}^3;d^3x)\big),$$ (cf. Simon [@Si71],_Theorem_I.22_and Example_4 in Sect. I.6)_and_hence $V_0$_is_$H_0$-form compact, but $${\text{\rm{dom}}}(V_0) \cap {\text{\rm{dom}}}(H_0)_=\{0\},$$_and thus $V_0$ is not $H_0

1], which appears to play an important role in combinatorial free probability (see review in Section 1 below). The formula $(I)$ can be extended to the case of an R-cyclic family of matrices (see Definition 2.9 and Theorem 2.10 in Section 2), and can be used to obtain various situations when one gets a family of matrices with computable joint distribution. Some applications are presented in the Section 3 of the paper. Section 4 is about operations with matrices in an R-cyclic family. It is trivial from the definition that if $A_{1}, \ldots, A_{s}$ is an R-cyclic family (of $d \times d$ matrices over a non-commutative probability space ${ ( {\cal A}, \varphi ) }$), then one can add to $A_{1}, \ldots, A_{s}$: \(a) a linear combination of $A_{1}, \ldots, A_{s}$, or \(b) any scalar diagonal matrix, and the enlarged family is still R-cyclic. In Lemma 4.2 we show that a similar statement is true when one adds to $A_{1}, \ldots, A_{s}$ a product of some of the matrices in the family; this comes as a fairly easy application of a formula for non-crossing cumulants with products for entries, which was found in [@KS]. The considerations of Section 4 show that the property of a family of matrices $A_{1}, \ldots, A_{s} \in M_{d} ( { {\cal A} })$ of being R-cyclic is really a property of the algebra ${\cal C}$ generated together by $A_{1}, \ldots, A_{s}$ and the set of scalar diagonal matrices. The rest of the paper is devoted to identifying what this property of ${\cal C}$ exactly is. The result turns out to be the following (Theorem 8.2): $$(II) \hspace{2.5cm} \left\{ \begin{array}{c} \mbox{the family} \\ A_{1}, \ldots, A_{s} \\ \mbox{is R-cyclic} \end{array} \right\} \ \Leftrightarrow \ \left\{ \begin{array}{c} \mbox{${\cal C}$ is free from $M_{d} ( { {\bf C} })$,} \\ \

1 ], which appears to play an important role in combinative spare probability (see review in incision 1 under). The formula $ (I)$ can be extended to the character of an gas constant - cyclic family of matrices (see Definition 2.9 and Theorem 2.10 in Section 2), and can be use to obtain various situation when one gets a family of matrices with computable joint distribution. Some applications are presented in the Section 3 of the paper. part 4 is about operations with matrices in an R - cyclic syndicate. It is trivial from the definition that if $ A_{1 }, \ldots, A_{s}$ is an R - cyclic family (of $ vitamin d \times d$ matrices over a non - commutative probability space $ { ({ \cal A }, \varphi) } $), then one can add to $ A_{1 }, \ldots, A_{s}$: \(a) a linear combination of $ A_{1 }, \ldots, A_{s}$, or \(b) any scalar diagonal matrix, and the exaggerated family is still R - cyclic. In Lemma 4.2 we show that a similar statement is true when one adds to $ A_{1 }, \ldots, A_{s}$ a product of some of the matrices in the family; this comes as a fairly easy application of a formula for non - crossing cumulants with products for entry, which was found in [ @KS ]. The considerations of incision 4 show that the property of a family of matrices $ A_{1 }, \ldots, A_{s } \in M_{d } ({ { \cal A } }) $ of being radius - cyclic is really a property of the algebra $ { \cal C}$ generated together by $ A_{1 }, \ldots, A_{s}$ and the hardening of scalar diagonal matrices. The rest of the paper is devoted to identifying what this place of $ { \cal C}$ exactly is. The result turns out to be the following (Theorem 8.2 ): $ $ (II) \hspace{2.5 cm } \left\ { \begin{array}{c } \mbox{the family } \\ A_{1 }, \ldots, A_{s } \\ \mbox{is R - cyclic } \end{array } \right\ } \ \Leftrightarrow \ \left\ { \begin{array}{c } \mbox{${\cal C}$ is complimentary from $ M_{d } ({ { \bf C } }) $, } \\ \

1], whlch appears to play an ioportant role iu combiiatoriam free pfobability (see review in Secvion 1 beliw). The formula $(I)$ can bd extendef to the cast of an R-cyclic fejily of matridcs (sez Vefinition 2.9 and Theorem 2.10 hn Section 2), ang zau be used to obtain various situatiogs when oje gets a famijy og matdpcts with computable joint distribufion. Soke applicatioms are presented in the Seftioj 3 of the paper. Sechion 4 is abiut jperations wigh matrices in an R-cycmic family. It is trivial from tfe deyinition thqt if $D_{1}, \ldots, A_{s}$ ms an G-cyclic familn (of $d \dimes d$ matrices over a ion-cimmutative probabilitb space ${ ( {\cal A}, \varpri ) }$), then oue can add to $A_{1}, \ldots, A_{w}$: \(a) a ninedr cunbivatjoi or $A_{1}, \ldlts, A_{s}$, or \(b) ang scalar diqgonal matrix, and tne vmlarged famimy is ftyll R-cyclic. In Lemma 4.2 we show that a sikilzr statement is true whwn one adds to $A_{1}, \ldotd, A_{s}$ a prjduct of some of the matrices in the family; this womes xs c fairut fasy application of a formula for non-crossing dukukants with probucts for entriex, ahosh was found kn [@KS]. Tks donsiderations of Dection 4 shoq that tht prolerty of a family of matricws $A_{1}, \ldots, A_{f} \in M_{d} ( { {\cal A} })$ of being R-cycuic os reslly a property of the clgebrz ${\cal C}$ genfrated tofdther by $A_{1}, \ldots, A_{s}$ atd the stg of scalar diagogal matrires. Tke rest uf tne papqr is devohed to identifying what hhis kroperdy of ${\cal F}$ exactly is. The result turns onv to be the fplnowpng (Theorzm 8.2): $$(II) \hspace{2.5cm} \lefe\{ \begin{array}{c} \kbox{the famiuy} \\ A_{1}, \ldota, A_{s} \\ \mbmx{is R-cyclis} \end{array} \ricjt\} \ \Leftrighvarrow \ \lest\{ \vegib{array}{c} \ocox{${\cal C}$ is frre from $M_{b} ( { {\bd C} })$,} \\ \

1], which appears to play an important combinatorial probability (see in Section 1 be to the case an R-cyclic family matrices (see Definition 2.9 and Theorem in Section 2), and can be used to obtain various situations when one a family of matrices with computable joint distribution. Some applications are presented in Section of paper. 4 is about operations with matrices in an R-cyclic family. It is trivial from the definition if $A_{1}, \ldots, A_{s}$ is an R-cyclic family $d \times d$ matrices a non-commutative probability space ${ {\cal \varphi ) then can to $A_{1}, \ldots, \(a) a linear combination of $A_{1}, \ldots, A_{s}$, or \(b) any scalar diagonal matrix, and the enlarged is still Lemma 4.2 show a statement is true adds to $A_{1}, \ldots, A_{s}$ a of the matrices in the family; this comes a fairly application of a formula for non-crossing with products for entries, which was found in The considerations of Section 4 show that the property of a family of matrices $A_{1}, \in M_{d} ( { A} })$ of R-cyclic really property the algebra C}$ generated together by $A_{1}, \ldots, A_{s}$ and the set of diagonal matrices. The rest of the paper is devoted to this of ${\cal C}$ is. The result turns to the following (Theorem 8.2): \left\{ \mbox{the \ldots, \\ R-cyclic} \end{array} \right\} \ \ \left\{ \begin{array}{c} \mbox{${\cal C}$ free from $M_{d} ( \

1], which appears to play an imporTant role in CombiNatOriAl Free ProbAbility (see reviEW in SEction 1 below). The formula $(I)$ Can be ExTEndeD To The caSe of an R-CYcLIC faMiLy Of mAtRIcEs (see defInition 2.9 And Theorem 2.10 In SEcTion 2), and can be USeD to obtain vAriOus situationS whEn one gEtS a fAMily oF maTriceS with cOMputabLe joint diStRIbutioN. some appLICaTionS are presented in thE seCTion 3 of the paper. sectioN 4 iS AbOUT opEraTions with mAtRices IN an R-cycLIc FAMIly. iT is trivial froM the definitIOn tHat if $A_{1}, \LdOts, a_{S}$ is an R-CycliC fAMilY (of $d \times d$ mAtriCes over a nOn-commUTative pRObabiliTy spacE ${ ( {\caL A}, \vArphI ) }$), ThEn One CaN Add TO $A_{1}, \LdoTS, A_{s}$: \(A) a linear CoMbInatiOn of $a_{1}, \LDOTs, A_{s}$, Or \(b) Any sCalar Diagonal matriX, anD the ENlaRged fAmily Is stIlL R-cycLic. In LEmma 4.2 wE sHow that a similar StatEment is trUe wHeN onE aDds to $a_{1}, \Ldots, A_{S}$ a pRodUct of soMe of the MAtrIcES IN tHe family; this comes aS a FAIrLy easy apPlicatIOn Of A Formula fOr Non-CrosSINg cumUlanTS wIth produCts for ENtRiEs, which WaS found In [@kS]. THe cOnsidERatiOns of SEction 4 shOw thaT The property of a FAmily of matricES $A_{1}, \LDOtS, a_{s} \in m_{d} ( { {\cAl A} })$ of being R-CyclIC is rEallY A pRopERty of The alGeBRa ${\CAl C}$ generated togetheR bY $A_{1}, \ldotS, A_{s}$ anD the set of scalAr diagonal MATRices. The Rest OF tHE paper is devoteD to idEntifying wHAt this prOpertY of ${\cal C}$ eXactly is. THE Result tuRns Out To bE thE FOlLowing (Theorem 8.2): $$(ii) \HspaCe{2.5Cm} \left\{ \bEgiN{array}{c} \MboX{thE faMilY} \\ A_{1}, \Ldots, A_{s} \\ \mbOx{is R-cycLiC} \eNd{ArRay} \Right\} \ \lEftrightArRow \ \LeFt\{ \bEgin{aRRay}{c} \mbOx{${\cal c}$ is fReE fROm $M_{D} ( { {\bf C} })$,} \\ \

1], which appears to playan importa nt ro leinco mbin ator ial free proba b ilit y (see review in Secti on 1be l ow). Th e for mula $( I )$ c anbe e xte nd e dto th e c ase ofan R-cycli c f am ily of matri c es (see Defi nit ion 2.9 andThe orem 2 .1 0 i n Sect ion 2),and ca n be us ed to obt ai n vario u s situa t i on s wh en one gets a fam i ly of matrices wi th com pu t ab l e jo int distribut io n. So m e appli c at i o n s a r e presented i n the Secti o n 3 of th epap e r. Se ction 4 isabout opera tion s with ma trices in an R - cyclicfamily . I t i s tr i vi al fr om the de fin i tio n that i f$A _{1}, \ld o t s , A_{ s}$ isan R- cyclic family (o f $d \ti mes d $ mat rice sovera non- commu ta tive probabilit y sp ace ${ ({\c al A} ,\varp h i ) }$ ),the n one c an addt o $ A_ { 1 } ,\ldots, A_{s}$: \ (a ) alinear c ombina t io no f $A_{1} ,\ld ots, A _{s}$ , or \ (b) anyscalar di ag onal ma tr ix, an dthe en large d fam ily is still R -cycl i c. In Lemma 4. 2 we show that as i mi l ar s tat ement is tr ue w h en o ne a d ds to $A_{1 }, \l do t s, A_{s}$ a product of s ome of thematrices in t he family; t h is comes asa f a irly easy appl icati on of a fo r mula for non- crossing cumulant s with pro duc tsfor en t r ie s, which wasf o undin [@KS]. T he cons ide rat ion s o fSection 4 show th at t he p rop ertyo f a fami ly of m atr ices$ A_{1}, \ldo ts,A_ {s } \i n M_{d} ({ {\ca lA} })$ of b eingR-cy c lic is rea lly a pro per t y of t he algebr a ${\cal C}$ge nerated to ge the r by $ A _ {1}, \ld ots, A_{s}$ and the set of scal ardiago nalmatrices. Th e rest of the pa per is devo te d t o ident i f yi ngwh at this pr o p ert y of${ \cal C}$ ex actly is. The resu l t t urns out to b e t he f o l lo win g ( T heo re m 8. 2 ) : $$(II) \hspac e{2.5cm} \l e ft \{ \begin { arr ay }{c} \m box{the fami l y} \\A_{1}, \l dots, A_{ s} \\\ m box {is R-cycl ic} \end {array} \ right \ }\ \Le ftr ightar ro w \ \lef t\{ \ b egi n{arr ay}{c} \ mbox{$ {\cal C }$ is fr ee from $M_{d} ( { {\bf C} }) $,} \\\

1], which_appears to_play an important role_in combinatorial_free_probability (see_review_in Section 1_below). The formula_$(I)$ can be extended_to the case_of_an R-cyclic family of matrices (see Definition 2.9 and Theorem 2.10 in Section 2),_and_can be_used_to_obtain various situations when one_gets a family of matrices_with computable_joint distribution. Some applications are presented in the_Section_3 of the_paper. Section 4 is about operations with matrices in an_R-cyclic family. It is trivial from_the definition that_if_$A_{1},_\ldots, A_{s}$ is an_R-cyclic family (of $d \times d$_matrices over a non-commutative probability space_${ ( {\cal A}, \varphi ) }$),_then one can add to $A_{1},_\ldots, A_{s}$: \(a) a linear combination_of $A_{1},_\ldots, A_{s}$, or \(b) any scalar_diagonal matrix, and_the enlarged_family is still_R-cyclic. In Lemma 4.2 we show_that a similar_statement is true when one adds_to_$A_{1}, \ldots, A_{s}$_a_product_of some_of the matrices_in_the family;_this_comes as a fairly easy application_of_a formula for non-crossing cumulants with products_for entries, which was_found_in [@KS]. The considerations of_Section 4 show that the_property of a family of matrices_$A_{1}, \ldots,_A_{s} \in_M_{d} ( { {\cal A} })$ of being R-cyclic is really_a property of the algebra ${\cal_C}$ generated together by_$A_{1}, \ldots,_A_{s}$_and the set_of_scalar diagonal_matrices. The rest of the paper is_devoted to_identifying what this property of ${\cal_C}$ exactly is. The_result_turns out to be the following_(Theorem 8.2): $$(II) \hspace{2.5cm} \left\{ _\begin{array}{c} \mbox{the family} \\ A_{1}, \ldots,_A_{s}_\\ \mbox{is_R-cyclic} \end{array} \right\} \ \Leftrightarrow_\ \left\{ \begin{array}{c} \mbox{${\cal C}$ is free_from $M_{d} (_{ {\bf C} })$,} \\ \

function. The case of a tensor-valued distribution function, even though it is the less trivial part, is treated in appendix \[AppGTtensor\]. We then infer what should be the transformation rule of the Liouville and collision operators, and in order to complete the consistency test, we check that the perturbed expressions of the Liouville and collision operators do indeed transform following these rules. Coordinates on the manifold --------------------------- We need to specify how the functional dependence of the quantities appearing in the Boltzmann equation (and in the Einstein equation) is obtained. If we consider a scalar function $f:{\cal M}\mapsto {\mathbb{R}}$ on the space-time manifold ${\cal M}$, then the choice of a coordinates system [^1] $c:{\mathbb{R}}^4 \mapsto {\cal M}$ does not affect the geometrical meaning of this function, but it affects its functional form $f \circ c:{\mathbb{R}}^4 \mapsto {\mathbb{R}}$, where $\circ$ designates the composition rule, in the sense that for another coordinates system $\tilde c:{\mathbb{R}}^4 \mapsto {\cal M}$, then $f\circ \tilde c \neq f \circ c$. The confusion only arises from the fact that we often refer to $f \circ c$ as $f$ only. A distribution function $f$ (that we take as a scalar-valued for simplicity here) is a function on the tangent bundle $T {\cal M}$ of the manifold and can be regarded as a function on manifold ${\cal M}$ describing the space-time and on the tangent space ${\mathbb{R}}^4$ (more precisely a restriction to the mass shell ${\mathbb{R}}^3$) of each point. It is thus a function $$f: (T{\cal M})\mapsto {\mathbb{R}}\,.$$ Again the particular choice of coordinates on the tangent space does not affect the geometrical meaning of the function but its functional form. Once a choice $c$ of coordinates on the manifold ${\cal M}$ has been made, there is a natural basis, called canonical basis, that is made of the partial derivatives with respect to the coordinates. This leads to a natural coordinates system $Tc$ for the tangent space at each point. Thus $(c,Tc) : {\mathbb{R}}^4 \times {\mathbb{R}}^3 \mapsto T{\cal M}$, is a coordinates system for the tangent bundle. In order to simplify the notation we will note $(c

function. The case of a tensor - valued distribution function, even though it is the less fiddling character, is treated in appendix   \[AppGTtensor\ ]. We then infer what should be the transformation rule of the Liouville and collision operator, and in order to complete the consistency examination, we discipline that the perturbed expression of the Liouville and collision operators do indeed transform keep up these rules. Coordinates on the manifold --------------------------- We want to intend how the functional dependence of the quantities appearing in the Boltzmann equation (and in the Einstein equality) is obtained. If we consider a scalar function $ f:{\cal M}\mapsto { \mathbb{R}}$ on the distance - time manifold $ { \cal M}$, then the choice of a coordinate system [ ^1 ] $ c:{\mathbb{R}}^4 \mapsto { \cal M}$ does not feign the geometrical meaning of this routine, but it affects its functional form $ f \circ c:{\mathbb{R}}^4 \mapsto { \mathbb{R}}$, where $ \circ$ delegate the composition rule, in the sense that for another coordinates system $ \tilde c:{\mathbb{R}}^4 \mapsto { \cal M}$, then $ f\circ \tilde c \neq f \circ c$. The confusion merely arises from the fact that we often refer to $ f \circ c$ as $ f$ only. A distribution function $ f$ (that we take as a scalar - valued for simplicity here) is a function on the tangent bundle $ T { \cal M}$ of the manifold paper and can be see as a function on manifold $ { \cal M}$ identify the quad - time and on the tangent space $ { \mathbb{R}}^4 $ (more precisely a restriction to the bulk shell $ { \mathbb{R}}^3 $) of each point. It is thus a function $ $ f: (T{\cal M})\mapsto { \mathbb{R}}\,.$$ Again the especial choice of coordinates on the tangent space does not affect the geometrical meaning of the function but its running form. Once a choice $ c$ of coordinates on the manifold $ { \cal M}$ has been made, there is a natural basis, call canonical basis, that is made of the partial derivatives with deference to the coordinates. This leads to a natural coordinates system $ Tc$ for the tangent space at each point. Thus $ (deoxycytidine monophosphate, Tc): { \mathbb{R}}^4 \times { \mathbb{R}}^3 \mapsto T{\cal M}$, is a coordinates system for the tangent bundle. In order to simplify the note we will note $ (c

fujction. The case of a tenror-valued distrnvution functjon, even though it is the less triviel pqrt, iw treated in appendix \[AopGTtensog\]. We then inftr what should be the trakfforjwtiou cule of the Liooville and cmllision operadofs, and in order to complete the consiftency yedt, we check thwt tnq pednuvbed expressions of the Liouvills and cmllision operstors do indeed transform volllwing these rules. Clordinates in trw manifold --------------------------- We need to skeeify how thg functional dependence of the qjanticies appearunt ij the Boltzmenn eqlation (and in the Einvtein eauation) is obtsinxd. Id we consider a scalac function $f:{\cal M}\mapfto {\mathbt{R}}$ on the space-time mqnifond ${\cdl M}$, rhev tge cgoice lf e coordinatss system [^1] $x:{\mathbb{R}}^4 \mapsto {\cal M}$ eoes not affedt the gqometrical meaning of this function, but it affects its functional form $f \circ c:{\mathbb{R}}^4 \mapsto {\mwthbb{R}}$, where $\circ$ designates the composition rule, in tie seuwe thxr vor another coordinates system $\tilde c:{\mathbb{R}}^4 \jakstp {\cal M}$, then $f\girc \tilde c \neq f \clrv c$. The confuskon only zrises from the faft that we odten refew to $f \circ c$ as $f$ only. A distrubution funcnion $f$ (that we take as a scalar-vauued for ximplicity here) is a fuuction on the tanhent bundmd $T {\cal M}$ of the makifmld and can be regarded as a functiin ou manifoud ${\csl M}$ dqscribing hhe sidce-time and on the tanggnt spdce ${\mathbb{G}}^4$ (more precisely a restriction vp the mass snenl ${\kathbb{R}}^3$) jf eagh point. It is ehus a functiou $$f: (T{\cal M})\maprto {\mathbb{G}}\,.$$ Again tie particulaw choice of cmlrdinates on the tandent spaxe does vot affect the geometribaj meqning of the functlon bot its functional fotn. Once a choicr $c$ of clocdinaeas on the matifoud ${\zsl M}$ fas been mabe, thete is a natural basiv, camled canonical basos, that is made of the partial cerivatives with rtspect to the vootdinates. This leads to a naturam coordinwtef system $Tc$ sor bhe jangent spaee at each point. Thus $(c,Tc) : {\mathbb{R}}^4 \times {\mathbb{R}}^3 \mapsto T{\cql M}$, is a coordinatgs xystem for vhe taggent bungle. In order to simpoify the notation we will note $(c

function. The case of a tensor-valued distribution though is the trivial part, is then what should be transformation rule of Liouville and collision operators, and in to complete the consistency test, we check that the perturbed expressions of the and collision operators do indeed transform following these rules. Coordinates on the manifold We to how functional dependence of the quantities appearing in the Boltzmann equation (and in the Einstein equation) is If we consider a scalar function $f:{\cal M}\mapsto on the space-time manifold M}$, then the choice of coordinates [^1] $c:{\mathbb{R}}^4 {\cal does affect the geometrical of this function, but it affects its functional form $f \circ c:{\mathbb{R}}^4 \mapsto {\mathbb{R}}$, where $\circ$ designates composition rule, sense that another system c:{\mathbb{R}}^4 \mapsto {\cal $f\circ \tilde c \neq f \circ only arises from the fact that we often to $f c$ as $f$ only. A distribution $f$ (that we take as a scalar-valued for here) is a function on the tangent bundle $T {\cal M}$ of the manifold and regarded as a function manifold ${\cal M}$ the and the space ${\mathbb{R}}^4$ precisely a restriction to the mass shell ${\mathbb{R}}^3$) of each point. is thus a function $$f: (T{\cal M})\mapsto {\mathbb{R}}\,.$$ Again the of on the tangent does not affect the meaning the function but its Once choice on manifold M}$ has been made, is a natural basis, called basis, that is made respect to the coordinates. This leads to a coordinates system $Tc$ for the tangent space each point. Thus $(c,Tc) : {\mathbb{R}}^4 \times {\mathbb{R}}^3 \mapsto T{\cal M}$, is coordinates system tangent bundle. In order to simplify the notation will note $(c

function. The case of a tensor-vAlued distrIbutiOn fUncTiOn, evEn thOugh it is the lesS TrivIal part, is treated in appeNdix \[APpgttenSOr\]. we theN infer wHAt SHOulD bE tHe tRaNSfOrmatIon Rule of tHe LiouvillE anD cOllision operAToRs, and in ordEr tO complete the ConSistenCy TesT, We cheCk tHat thE pertuRBed expRessions oF tHE LiouvILle and cOLLiSion Operators do indeed TRaNSform following These rUlES. COORdiNatEs on the manIfOld --------------------------- We NEed to spECiFY HOw tHE functional dePendence of tHE quAntitiEs AppEAring iN the BOlTZmaNn equation (aNd in The EinsteIn equaTIon) is obTAined. If We consIdeR a sCalaR FuNcTioN $f:{\CAl M}\MApSto {\MAthBb{R}}$ on the SpAcE-time ManiFOLD ${\Cal M}$, TheN the ChoicE of a coordinatEs sYsteM [^1] $C:{\maThbb{R}}^4 \MapstO {\cal m}$ dOes noT affecT the gEoMetrical meaning Of thIs functioN, buT iT afFeCts itS FunctiOnaL foRm $f \circ C:{\mathbb{r}}^4 \MapStO {\MAThBb{R}}$, where $\circ$ designAtES ThE composiTion ruLE, iN tHE sense thAt For AnotHER coorDinaTEs System $\tiLde c:{\maTHbB{R}}^4 \Mapsto {\cAl m}$, then $f\CiRc \tIldE c \neq F \Circ C$. The coNfusion oNly arISes from the fact THat we often refER tO $F \CiRC c$ as $F$ onLy. A distribuTion FUnctIon $f$ (THaT we TAke as A scalAr-VAlUEd for simplicity here) Is A functIon on The tangent bunDle $T {\cal M}$ of THE Manifold And cAN bE Regarded as a funCtion On manifold ${\CAl M}$ descrIbing The space-Time and on THE tangent SpaCe ${\mAthBb{R}}^4$ (MORe Precisely a resTRIctiOn To the maSs sHell ${\matHbb{r}}^3$) of EacH poInT. It is thus A functioN $$f: (t{\cAl m})\mApsTo {\matHBb{R}}\,.$$ Again ThE paRtIcuLar chOIce of cOordiNateS oN tHE taNgent spACe DOEs noT aFfEct tHe gEoMetriCal mEAniNg of the Function bUt iTS funCtIoNal form. once a choice $c$ oF cOordinates On The ManifoLD ${\Cal M}$ has bEen made, there is a natural bASis, callEd cAnoniCal bAsis, that iS maDe of thE paRTial deRivatiVes wiTh ResPECt to tHE CoOrdInAtes. This leADS to A natuRaL cooRdinateS system $Tc$ for the tanGEnt Space at each poInt. thus $(C,tC) : {\mAthBB{R}}^4 \TImeS {\mAThbB{r}}^3 \Mapsto T{\cal M}$, is a cOordinates SySTeM for the tanGEnt BuNdle. In oRder to sImpliFY the notAtion we wiLl note $(c

function. The case of a t ensor-valu ed di str ibu ti on f unct ion, even thou g h it is the less trivial p art,is trea t ed in a ppendix \[ A p pGT te ns or\ ]. We then in fer wha t should b e t he transformat i on rule of t heLiouville an d c ollisi on op e rator s,and i n orde r to co mplete th ec onsist e ncy tes t , w e ch eck that the pert u rb e d expressionsof the L i ou v i lle an d collisio nopera t ors doi nd e e d tr a nsform follow ing these r u les . Coo rd ina t es onthe m an i fol d --------- ---- --------- -----We need to spec ify ho w t hefunc t io na l d ep e nde n ce of the quantit ie sappea ring i n theBol tzma nn eq uation (and i n t he E i nst ein e quati on)is obta ined.If we c onsider a scala r fu nction $f :{\ ca l M }\ mapst o {\mat hbb {R} }$ on t he spac e -ti me m a ni fold ${\cal M}$, t he n th e choice of ac oo rd i nates sy st em[^1] $ c:{\m athb b {R }}^4 \ma psto { \ ca lM}$ doe snot af fe ctthe geom e tric al mea ning ofthisf unction, but i t affects itsf un c t io n al f orm $f \circ c :{\m a thbb {R}} ^ 4\ma p sto { \math bb { R} } $, where $\circ$ de si gnates thecomposition r ule, in th e s ense tha t fo r a n other coordina tes s ystem $\ti l de c:{\m athbb {R}}^4 \ mapsto {\ c a l M}$, t hen $f \ci rc\ t il de c \neq f \ c i rc c $. The co nfu sion on lyari ses fr om the fact that we o ft en r efe r to$ f \circc$ as $ f$only. A dist ribut ionfu nc t ion $f$ (t h at w e ta ke a s asca la r-val uedf orsimplic ity here) is a fu nc ti on on t he tangent bu nd le $T {\ca lM}$ of th e manifold and can be regarded as a funct ion on m anif old ${\ca l M }$ des cri b ing th e spac e-tim eand o n the t an gen tspace ${\m a t hbb {R}}^ 4$ (mo re prec isely a restrictio n to the mass she ll${\m a t hb b{R } }^ 3 $)of eac h point. It is th us a funct io n $ $f: (T{\ca l M} )\ mapsto{\mathb b{R}} \ ,.$$ Ag ain the p articular c hoic e ofcoordinate s on the tangents paced oe s not af fect t he ge ometr ical m e ani ng of the f un ctionbut i ts functio nal form. Once a choice $c$ o f coo rdi nates onthe man ifold ${\ calM}$ has be enmad e, th ere is anatu r al ba s is, c alle d canonica l b asi s , t hat is made o f th e par tia l deriv ativ es with respect t o the coordinat es.T h islea d s to a natural coord ina te s system $ Tc $ for the t angent s pa c e ateach p oint.Thus $( c , Tc ) : {\m athb b{R }}^4 \tim es{\ m athbb{R }} ^3 \mapst o T{ \c al M}$ , is a coor d i nates system for thet a ngent bun dle.In ordert o si mplify the notation w e will not e $(c

function._The case_of a tensor-valued distribution_function, even_though_it is_the_less trivial part,_is treated in_appendix \[AppGTtensor\]. We then infer_what should be_the_transformation rule of the Liouville and collision operators, and in order to complete the_consistency_test, we_check_that_the perturbed expressions of the_Liouville and collision operators do_indeed transform_following these rules. Coordinates on the manifold --------------------------- We need to_specify_how the functional_dependence of the quantities appearing in the Boltzmann equation_(and in the Einstein equation) is_obtained. If we_consider_a_scalar function $f:{\cal M}\mapsto_{\mathbb{R}}$ on the space-time manifold ${\cal_M}$, then the choice of a_coordinates system [^1] $c:{\mathbb{R}}^4 \mapsto {\cal M}$_does not affect the geometrical meaning_of this function, but it_affects its_functional form $f \circ c:{\mathbb{R}}^4_\mapsto {\mathbb{R}}$, where_$\circ$ designates_the composition rule,_in the sense that for another_coordinates system $\tilde_c:{\mathbb{R}}^4 \mapsto {\cal M}$, then $f\circ_\tilde_c \neq f_\circ_c$._The confusion_only arises from_the_fact that_we_often refer to $f \circ c$_as_$f$ only. A distribution function $f$ (that_we take as a_scalar-valued_for simplicity here) is_a function on the tangent_bundle $T {\cal M}$ of the_manifold and_can be_regarded as a function on manifold ${\cal M}$ describing the space-time_and on the tangent space ${\mathbb{R}}^4$_(more precisely a restriction_to the_mass_shell ${\mathbb{R}}^3$) of_each_point. It_is thus a function $$f: (T{\cal M})\mapsto_{\mathbb{R}}\,.$$ Again_the particular choice of coordinates on_the tangent space does_not_affect the geometrical meaning of the_function but its functional form. Once_a choice $c$ of coordinates_on_the_manifold ${\cal M}$ has been_made, there is a natural basis,_called canonical basis,_that is made of the partial derivatives_with_respect to the coordinates. This leads_to_a natural coordinates system $Tc$ for_the_tangent_space at each point. Thus_$(c,Tc) : {\mathbb{R}}^4 \times {\mathbb{R}}^3 \mapsto_T{\cal M}$, is a coordinates system for the tangent_bundle. In order_to simplify the notation we_will_note_$(c

models based on Hawkes processes for capturing reciprocity and stochastic block-models or latent feature models for capturing community structure. In addition to the above two properties, it is important to capture the global properties of the interaction data. Interaction data are often *sparse*: only a small fraction of the pairs of nodes actually interact. Additionally, they typically exhibit high degree (number of interactions per node) *heterogeneity*: some individuals have a large number of interactions, whereas most individuals have very few, therefore resulting in empirical degree distributions being heavy-tailed. As shown by @Karrer2011, @Gopalan2013 and @Todeschini2016, failing to account explicitly for degree heterogeneity in the model can have devastating consequences on the estimation of the latent structure. Recently, two classes of statistical models, based on random measures, have been proposed to capture sparsity and power-law degree distribution in network data. The first one is the class of models based on exchangeable random measures [@Caron2017; @Veitch2015; @Herlau2015; @Borgs2016; @Todeschini2016; @Palla2016; @Janson2017]. The second one is the class of edge-exchangeable models [@Crane2015; @Crane2017; @Cai2016; @Williamson2016; @Janson2017a; @Ng2017]. Both classes of models can handle both sparse and dense networks and, although the two constructions are different, connections have been highlighted between the two approaches [@Cai2016; @Janson2017a]. The objective of this paper is to propose a class of statistical models for temporal dyadic interaction data that can capture all the desired properties mentioned above, which are often found in real world interactions. These are *sparsity*, *degree heterogeneity*, *community structure* and *reciprocity*. Combining all the properties in a single model is non trivial and there is no such construction to our knowledge. The proposed model generalises existing reciprocating relationships models [@Blundell2012] to the sparse and power-law regime. Our model can also be seen as a natural extension of the classes of models based on exchangeable random measures and edge-exchangeable models and it shares properties of both families. The approach is shown to outperform alternative models for link prediction on a variety of temporal network datasets. The construction is based on Hawkes processes and the (static) model of @Todeschini2016 for sparse and modular graphs

models based on Hawkes processes for capturing reciprocality and stochastic blockage - models or latent feature model for capture community structure. In summation to the above two property, it is important to get the global properties of the interaction datum. Interaction data are often * sparse *: only a small fraction of the pair of nodes actually interact. Additionally, they typically exhibit high degree (number of interaction per node) * heterogeneity *: some individuals have a big number of interactions, whereas most individuals have very few, consequently resulting in empiric degree distributions being heavy - tailed. As shown by @Karrer2011, @Gopalan2013 and @Todeschini2016, fail to account explicitly for degree heterogeneity in the model can have devastating consequence on the estimation of the latent structure. Recently, two classes of statistical models, based on random measures, have been proposed to capture sparsity and power - jurisprudence degree distribution in network datum. The inaugural one is the course of models based on exchangeable random measures   [ @Caron2017; @Veitch2015; @Herlau2015; @Borgs2016; @Todeschini2016; @Palla2016; @Janson2017 ]. The second one is the class of edge - exchangeable models   [ @Crane2015; @Crane2017; @Cai2016; @Williamson2016; @Janson2017a; @Ng2017 ]. Both class of models can handle both sparse and dense networks and, although the two constructions are different, connections have been highlight between the two approaches   [ @Cai2016; @Janson2017a ]. The objective of this paper is to propose a course of statistical models for worldly dyadic interaction data that can capture all the desired properties mentioned above, which are often determine in real world interactions. These are * sparseness *, * degree heterogeneity *, * residential district structure * and * reciprocity *. Combining all the properties in a single mannequin is non trivial and there embody no such construction to our cognition. The proposed model generalises existing reciprocate relationships models [ @Blundell2012 ] to the sparse and power - law regime. Our mannequin can also be seen as a natural extension of the classes of models based on exchangeable random measures and boundary - exchangeable models and it share properties of both families. The approach is shown to outperform alternative models for radio link prediction on a variety of temporal network datasets. The construction is based on Hawkes processes and the (inactive) exemplar of @Todeschini2016 for sparse and modular graph

mofels based on Hawkes progesses for captutibg recmprocitg and stuchastic block-models or lateit fwaturt models for capturkng commujity stryctuce. In addition to the aboyz two iropextmes, it is imporjant to captgre the global pfo'erties of the interaction data. Intewaction dwta are often *fparxq*: onmj c small fraction of the pairs or nodes actually intrract. Additionally, they tyoicaply exhibit high dfgree (numbet of unteractions per node) *heterogeneiti*: some individuals have a large vumbex of interaxtuond, whereas mowt inqividuals hayv very xew, thetefore resultikg in empirical degree distriuutions being heavy-twiled. As vhkwn by @Karrer2011, @Gopqlqn2013 ang @Togescfuni2016, fajlmng to acfouit explicitmy for degrwe heterogeneity in trv model can hzve dedaftating consequences on the estimation mf fhe latent structure. Recwntly, two classes of dtatisticwl models, based on random measures, have been propmsed vo caktmve soqrdity and power-law degree distribution in netwjdk dsta. The first jne is the vlwsx of models bared on exdhangeable random leasuref [@Carob2017; @Veitch2015; @Rerlsu2015; @Borgs2016; @Todeschini2016; @Palla2016; @Jqnson2017]. The sebond one is the class lf edge-exchcngeabke mocels [@Crane2015; @Crane2017; @Cai2016; @Wiluiamaon2016; @Janson2017a; @Ng2017]. Both duasses of models cak hdndle bouf sparse and densq networkw anb, althoueh tne two constructlons are different, conneftiond vave been jighlighted between the two appcpaches [@Cai2016; @Jamsmn2017a]. Nhe objeccive on this paper is to propose a elass of statkstical mosels foc temporal dradic interacdlon data thav can capeure all the derkred propertiex mentionvd aboce, which are often foons in real world inueeactions. These ard *s[agsivy*, *dedsee heterogeteith*, *cukmunigy structurc* avd *rrciprocity*. Combining all the properties in a single nodel is non trivial snd there is no sufh coistrucvion tp oor knowledge. The proposed model generalides existing resipricating relacionships models [@Blundell2012] to the sparse end power-law regime. Out model can also be sgen as a naturel extqnsion of the classes of modeos based on exchakgeable random measurea and adge-edchangeable models and it shares properties of both families. The approach is shown to outpedfork altvrnctire modejs flc link prediction on a variety of temporal netwock datasetv. Tke construction is based on Hswyes processes and the (static) model of @Todeschoni2016 for sparse and modular grapns

models based on Hawkes processes for capturing stochastic or latent models for capturing the two properties, it important to capture global properties of the interaction data. data are often *sparse*: only a small fraction of the pairs of nodes interact. Additionally, they typically exhibit high degree (number of interactions per node) *heterogeneity*: individuals a number interactions, whereas most individuals have very few, therefore resulting in empirical degree distributions being heavy-tailed. As by @Karrer2011, @Gopalan2013 and @Todeschini2016, failing to account for degree heterogeneity in model can have devastating consequences the of the structure. two of statistical models, on random measures, have been proposed to capture sparsity and power-law degree distribution in network data. The one is of models on random [@Caron2017; @Veitch2015; @Herlau2015; @Palla2016; @Janson2017]. The second one is edge-exchangeable models [@Crane2015; @Crane2017; @Cai2016; @Williamson2016; @Janson2017a; @Ng2017]. classes of can handle both sparse and dense and, although the two constructions are different, connections been highlighted between the two approaches [@Cai2016; @Janson2017a]. The objective of this paper is to class of statistical models temporal dyadic interaction that capture the properties mentioned which are often found in real world interactions. These are *sparsity*, heterogeneity*, *community structure* and *reciprocity*. Combining all the properties in model non trivial and is no such construction our The proposed model generalises relationships [@Blundell2012] and regime. model can also be as a natural extension of classes of models based edge-exchangeable models and it shares properties of both The approach is shown to outperform alternative for link prediction on a variety of temporal network datasets. The construction based on and the (static) model of @Todeschini2016 for sparse modular graphs

models based on Hawkes procesSes for captUring RecIprOcIty aNd stOchastic block-mODels Or latent feature models fOr capTuRIng cOMmUnity StructuRE. IN ADdiTiOn To tHe ABoVe two ProPerties, It is importAnt To Capture the glOBaL propertieS of The interactiOn dAta. IntErActIOn datA arE ofteN *sparsE*: Only a sMall fractIoN Of the pAIrs of noDES aCtuaLly interact. AdditiONaLLy, they typicallY exhibIt HIgH DEgrEe (nUmber of intErActioNS per nodE) *HeTEROgeNEity*: some indivIduals have a LArgE numbeR oF inTEractiOns, whErEAs mOst individuAls hAve very feW, thereFOre resuLTing in eMpiricAl dEgrEe diSTrIbUtiOnS BeiNG hEavY-TaiLed. As shoWn By @karreR2011, @GopALAN2013 And @TOdeSchiNi2016, faiLing to account ExpLiciTLy fOr degRee heTeroGeNeity In the mOdel cAn Have devastating ConsEquences oN thE eStiMaTion oF The latEnt StrUcture. REcently, TWo cLaSSES oF statistical models, BaSED oN random mEasureS, HaVe BEen propoSeD to CaptURE sparSity ANd Power-law Degree DIsTrIbution In NetworK dAta. the First ONe is The claSs of modeLs basED on exchangeablE Random measureS [@caRON2017; @VEItch2015; @herLau2015; @Borgs2016; @TodEschINi2016; @PaLla2016; @JANsOn2017]. THE secoNd one Is THe CLass of edge-exchangeaBlE modelS [@CranE2015; @Crane2017; @Cai2016; @WillIamson2016; @JansON2017A; @ng2017]. Both clAsseS Of MOdels can handle Both sParse and deNSe networKs and, Although The two conSTRuctions Are DifFerEnt, CONnEctions have beEN HighLiGhted beTweEn the twO apProAchEs [@CAi2016; @janson2017a]. ThE objectiVe Of ThIs PapEr is tO Propose a ClAss Of StaTistiCAl modeLs for TempOrAl DYadIc interACtION datA tHaT can CapTuRe all The dESirEd propeRties mentIonED aboVe, WhIch are oFten found in reAl World interAcTioNs. ThesE ARe *sparsiTy*, *degree heterogeneity*, *coMMunity sTruCture* And *rEciprocitY*. CoMbininG alL The proPertieS in a sInGle MODel is NON tRivIaL and there iS NO suCh conStRuctIon to ouR knowledge. The propoSEd mOdel generalisEs eXistING rEciPRoCAtiNg RElaTIOnships models [@BlUndell2012] to thE sPArSe and power-LAw rEgIme. Our mOdel can Also bE Seen as a Natural exTension of ThE claSSEs oF models basEd on exchAngeable rANdom mEAsUres aNd eDge-excHaNgeAble mOdels aND it ShareS propeRtIes of bOth faMiLies. The aPproach is shown to outperfOrm altErnatIve Models for LinK PreDiction on A varIety of tempOraL neTwork DatASets. THe coNStRucTIon is BaseD On Hawkes pROcEssES AnD the (static) mODEL of @todesChiNI2016 for spArse And modular graphs

models based on Hawkes pr ocesses fo r cap tur ing r ecip roci ty and stochas t ic b lock-models or latentfeatu re mode l sfor c apturin g c o m mun it ystr uc t ur e. I n a ddition to the ab ove t wo propertie s ,it is impo rta nt to captur e t he glo ba l p r opert ies of t he int e ractio n data. I nt e ractio n data a r e o ften *sparse*: only a sm a ll fraction of the p ai r so f no des actuallyin terac t . Addit i on a l l y,t hey typically exhibit hi g h d egree(n umb e r of i ntera ct i ons per node)*het erogeneit y*: so m e indiv i duals h ave alar genumb e rof in te r act i on s,w her eas most i nd ividu alsh a v e ver y f ew,there fore resultin g i n em p iri cal d egree dis tr ibuti ons be ing h ea vy-tailed. As s hown by @Karr er2 01 1,@G opala n 2013 a nd@To deschin i2016,f ail in g t oaccount explicitly f o r d egree he teroge n ei ty in the m od elcanh a ve de vast a ti ng conse quence s o nthe est im ationof th e l atent stru cture. Recent ly, t w o classes of s t atistical mod e ls , ba s ed o n r andom measu res, have bee n p rop o sed t o cap tu r es parsity and power-l aw degre e dis tribution innetwork da t a . The fir st o n ei s the class of mode ls based o n exchang eable randommeasures[ @ Caron201 7;@Ve itc h20 1 5 ;@Herlau2015;@ B orgs 20 16; @To des chini20 16; @P all a20 16 ; @Janson 2017]. T he s ec on d o ne is the clas sofed ge- excha n geable mode ls [ @C ra n e20 15; @Cr a ne 2 0 17;@C ai 2016 ; @ Wi lliam son2 0 16; @Janso n2017a; @ Ng2 0 17]. B ot h class es of modelsca n handle b ot h s parsea n d densenetworks and, althought he twocon struc tion s are dif fer ent, c onn e ctions havebeenhi ghl i g htedb e tw een t he two app r o ach es [@ Ca i201 6; @Jan son2017a]. The ob j ect ive of this p ape r is t opro p os e acl a sso f statistical mo dels for t em p or al dyadici nte ra ction d ata tha t can capture all thedesired p ro pert i e s m entioned a bove, wh ich are o f ten f o un d inrea l worl dint eract ions.T hes e are *spar si ty*, * degre eheteroge neity*, *community stru cture* and*re ciprocity *.C omb ining all the propertie s i n a sing lem odelis n o ntri v ial a nd t h ere is no su chc o ns truction to o u r k nowle dge . The p ropo sed model general i ses existing r ecip r o cat ing rela ti onships models [@ Bl u n dell2012 ]to the spar se and p ow e r-law regim e. Our modelc a na lso be see n a s a natur alex t ensionof t h e clas sesof model s base d one x changeable rando m mea s u res a n d e dge-e xc hangeab l e mo dels and i t shares pr operti es o f bot h famil ie s. The ap pr oach is sh o wn to out perfo rm alte rn ativ e m odelsforl i nk pr edic ti onon a vari e t yo fte m por al n etwor kdata sets. Th e constru cti o n is ba se d o n Hawkes pr o c esses andthe (sta t i c) model o f @To d es c hini2 016 fo r spar se andm odu la r graph s

models_based on_Hawkes processes for capturing_reciprocity and_stochastic_block-models or_latent_feature models for_capturing community structure. In_addition to the above_two properties, it_is_important to capture the global properties of the interaction data. Interaction data are often_*sparse*:_only a_small_fraction_of the pairs of nodes_actually interact. Additionally, they typically_exhibit high_degree (number of interactions per node) *heterogeneity*: some_individuals_have a large_number of interactions, whereas most individuals have very few,_therefore resulting in empirical degree distributions_being heavy-tailed. As_shown_by_@Karrer2011, @Gopalan2013 and @Todeschini2016,_failing to account explicitly for degree_heterogeneity in the model can have_devastating consequences on the estimation of the_latent structure. Recently, two classes of statistical_models, based on random measures,_have been_proposed to capture sparsity and_power-law degree distribution_in network_data. The first_one is the class of models_based on exchangeable_random measures [@Caron2017; @Veitch2015; @Herlau2015; @Borgs2016; @Todeschini2016;_@Palla2016;_@Janson2017]. The second_one_is_the class_of edge-exchangeable models [@Crane2015;_@Crane2017;_@Cai2016; @Williamson2016;_@Janson2017a;_@Ng2017]. Both classes of models can_handle_both sparse and dense networks and, although_the two constructions are_different,_connections have been highlighted_between the two approaches [@Cai2016; @Janson2017a]. The_objective of this paper is to_propose a_class of_statistical models for temporal dyadic interaction data that can capture all_the desired properties mentioned above, which_are often found in_real world_interactions._These are *sparsity*,_*degree_heterogeneity*, *community_structure* and *reciprocity*. Combining all the properties_in a_single model is non trivial and_there is no such_construction_to our knowledge. The proposed model_generalises existing reciprocating relationships models [@Blundell2012]_to the sparse and power-law_regime._Our_model can also be seen_as a natural extension of the_classes of models_based on exchangeable random measures and edge-exchangeable_models_and it shares properties of both_families._The approach is shown to outperform_alternative_models_for link prediction on a_variety of temporal network datasets. The construction_is based on Hawkes processes and the (static) model_of @Todeschini2016 for_sparse and modular graphs

. \[split-centre\] Let $G=H\rtimes\langle X\rangle$ be an abelian-by-procyclic group. Then $G$ has split centre if and only if $(G,G)\cap Z(G)=1$. It is clear that if $G$ has split centre then $(G,G)\cap Z(G)=1$. Conversely, suppose that $(G,G)\cap Z(G)=1$, and consider the $\mathbb{Z}_p$-module homomorphism $H\to H,h\mapsto (X,h)$.\ The kernel of this map is precisely $Z(G)$, therefore $(X,H)\cong\frac{H}{Z(G)}$. So since $Z(G)\cap (X,H)=1$, it follows that $Z(G)\times (X,H)$ has the same rank as $H$, hence it is open in $H$.\ Recall from [@Billy Definition 1.6] the definition of the *isolator* $i_G(N)$ of a closed, normal subgroup $N$ of $G$, and recall from [@Billy Proposition 1.7, Lemma 1.8] that it is a closed, isolated normal subgroup of $G$, and that $N$ is open in $i_G(N)$.\ Let $C=i_G((X,H))\leq H$, then it is clear that $Z(G)\cap C=1$ and that $Z(G)\times C$ is isolated. Therefore, since $Z(G)\times (X,H)$ is open in $H$, it follows that $Z(G)\times C=H$, and hence $G=Z(G)\times C\rtimes\langle X\rangle$, and $G$ has split centre. *Proof of Theorem \[B\].* Let $Z_1(G):=\{g\in G:(g,G)\subseteq Z(G)\}$, this is a closed subgroup of $G$ containing $Z(G)$ and contained in $H$. Suppose first that $Z_1(G)\neq Z(G)$.\ Then choose $h\in Z_1(G)\backslash Z(G)$, then $(h,G)\subseteq Z(G)$, so if we take $\psi\in Inn(G)$ to be conjugation by $h$, then

. \[split - centre\ ] Let $ G = H\rtimes\langle X\rangle$ be an abelian - by - procyclic group. Then $ G$ has split centre if and merely if $ (deoxyguanosine monophosphate, G)\cap Z(G)=1$. It is clear that if $ G$ has split centre then $ (guanine, G)\cap Z(G)=1$. Conversely, suppose that $ (G, G)\cap Z(G)=1 $, and regard the $ \mathbb{Z}_p$-module homomorphism $ H\to H, h\mapsto (X, h)$.\ The kernel of this function is precisely $ Z(G)$, therefore $ (X, H)\cong\frac{H}{Z(G)}$. So since $ Z(G)\cap (X, H)=1 $, it follows that $ Z(G)\times (X, H)$ has the same membership as $ H$, hence it is open in $ H$.\ Recall from [ @Billy Definition 1.6 ] the definition of the * isolator * $ i_G(N)$ of a shut, normal subgroup $ N$ of $ G$, and recall from [ @Billy Proposition 1.7, Lemma 1.8 ] that it is a closed, sequester normal subgroup of $ G$, and that $ N$ is open in $ i_G(N)$.\ Let $ coke = i_G((X, H))\leq H$, then it is clear that $ Z(G)\cap C=1 $ and that $ Z(G)\times C$ is isolated. Therefore, since $ Z(G)\times (X, H)$ is receptive in $ H$, it follows that $ Z(G)\times C = H$, and hence $ G = Z(G)\times C\rtimes\langle X\rangle$, and $ G$ has divide centre. * Proof of Theorem \[B\ ]. * Let $ Z_1(G):=\{g\in G:(g, G)\subseteq Z(G)\}$, this is a closed subgroup of $ G$ containing $ Z(G)$ and hold in $ H$. Suppose first that $ Z_1(G)\neq Z(G)$.\ Then choose $ h\in Z_1(G)\backslash Z(G)$, then $ (h, G)\subseteq Z(G)$, so if we take $ \psi\in Inn(G)$ to be conjugation by $ h$, then

. \[spllt-centre\] Let $G=H\rtimes\langue X\rangle$ be au abelien-by-prodyclic gfoup. Then $G$ has split centre id and only if $(G,G)\cap Z(G)=1$. It is clear thwt if $G$ yas wplit centrx then $(G,N)\eap Z(F)=1$. Conrecsely, suppose tmat $(G,G)\cap Z(C)=1$, and consider tfe $\mathbb{Z}_p$-module homomorphism $H\to H,h\iapsto (C,h)$.\ Hhe kernel of jhis kwp ia precisely $Z(G)$, therefore $(X,H)\cong\frad{H}{Z(G)}$. So since $Z(G)\cap (C,H)=1$, it follows that $Z(G)\times (X,H)$ jas the same rank ws $H$, hence ut if open in $H$.\ Rezall from [@Billy Definijion 1.6] the definition of the *isolxtor* $n_G(N)$ of a cliswd, jmrmal subgriup $N$ of $G$, and regsll frmm [@Billu Proposition 1.7, Lekma 1.8] that it is a closed, isolated normal sobgroup of $G$, and that $N$ is opwn in $i_C(N)$.\ Led $C=i_E((Z,H))\ldq G$, vheh it id cmear that $A(G)\cap C=1$ and that $Z(G)\times C$ is osjoated. Therefors, sincq $S(G)\times (X,H)$ is open in $H$, it follows that $Z(G)\fimes C=H$, and hence $G=Z(G)\tumes C\rtimes\langle X\rwngle$, and $G$ has split centre. *Proof of Theorem \[B\].* Let $Z_1(G):=\{g\in G:(c,G)\subadtew D(G)\}$, tfus is a closed subgroup of $G$ containing $Z(G)$ and sknuaiked in $H$. Suppose first that $Z_1(H)\nrz Z(G)$.\ Then choore $h\in Z_1(G)\gackslash Z(G)$, then $(j,G)\subsejeq Z(G)$, so if we takr $\psi\in Inn(G)$ to be conjugatuon by $h$, theu

. \[split-centre\] Let $G=H\rtimes\langle X\rangle$ be an Then has split if and only clear if $G$ has centre then $(G,G)\cap Conversely, suppose that $(G,G)\cap Z(G)=1$, and the $\mathbb{Z}_p$-module homomorphism $H\to H,h\mapsto (X,h)$.\ The kernel of this map is precisely therefore $(X,H)\cong\frac{H}{Z(G)}$. So since $Z(G)\cap (X,H)=1$, it follows that $Z(G)\times (X,H)$ has the rank $H$, it open in $H$.\ Recall from [@Billy Definition 1.6] the definition of the *isolator* $i_G(N)$ of a normal subgroup $N$ of $G$, and recall from Proposition 1.7, Lemma 1.8] it is a closed, isolated subgroup $G$, and $N$ open $i_G(N)$.\ Let $C=i_G((X,H))\leq then it is clear that $Z(G)\cap C=1$ and that $Z(G)\times C$ is isolated. Therefore, since $Z(G)\times (X,H)$ open in follows that C=H$, hence C\rtimes\langle X\rangle$, and split centre. *Proof of Theorem \[B\].* Z(G)\}$, this is a closed subgroup of $G$ $Z(G)$ and in $H$. Suppose first that $Z_1(G)\neq Then choose $h\in Z_1(G)\backslash Z(G)$, then $(h,G)\subseteq Z(G)$, if we take $\psi\in Inn(G)$ to be conjugation by $h$, then

. \[split-centre\] Let $G=H\rtimes\lanGle X\rangle$ Be an aBelIan-By-ProcYcliC group. Then $G$ has SPlit Centre if and only if $(G,G)\cap z(G)=1$. It iS cLEar tHAt If $G$ haS split cENtRE TheN $(G,g)\cAp Z(g)=1$. CONvErselY, suPpose thAt $(G,G)\cap Z(G)=1$, aNd cOnSider the $\mathBB{Z}_P$-module homOmoRphism $H\to H,h\mApsTo (X,h)$.\ ThE kErnEL of thIs mAp is pReciseLY $Z(G)$, theRefore $(X,H)\cOnG\Frac{H}{Z(g)}$. so since $z(g)\CaP (X,H)=1$, iT follows that $Z(G)\timES (X,h)$ Has the same rank As $H$, henCe IT iS OPen In $H$.\ recall from [@biLly DeFInition 1.6] THe DEFIniTIon of the *isolaTor* $i_G(N)$ of a clOSed, Normal SuBgrOUp $N$ of $G$, And reCaLL frOm [@Billy PropOsitIon 1.7, Lemma 1.8] tHat it iS A closed, ISolated Normal SubGroUp of $g$, AnD tHat $n$ iS OpeN In $I_G(N)$.\ lEt $C=I_G((X,H))\leq H$, ThEn It is cLear THAT $z(G)\caP C=1$ aNd thAt $Z(G)\tImes C$ is isolatEd. THereFOre, Since $z(G)\timEs (X,H)$ Is Open iN $H$, it foLlows ThAt $Z(G)\times C=H$, and hEnce $g=Z(G)\times C\RtiMeS\laNgLe X\raNGle$, and $g$ haS spLit centRe. *Proof OF ThEoREM \[b\].* LEt $Z_1(G):=\{g\in G:(g,G)\subseteq z(G)\}$, THIs Is a closeD subgrOUp Of $g$ ContainiNg $z(G)$ aNd coNTAined In $H$. SUPpOse first That $Z_1(G)\NEq z(G)$.\ then choOsE $h\in Z_1(G)\BaCksLasH Z(G)$, thEN $(h,G)\sUbseteQ Z(G)$, so if wE take $\PSi\in Inn(G)$ to be coNJugation by $h$, thEN

. \[split-centre\] Let $ G=H\rtimes \lang leX\r an gle$ bean abelian-by- p rocy clic group. Then $G$ h as sp li t cen t re if a nd only if $ (G, G) \c apZ( G )= 1$. Itis clea r that if$G$ h as split cen t re then $(G, G)\ cap Z(G)=1$. Co nverse ly , s u ppose th at $( G,G)\c a p Z(G) =1$, andco n sidert he $\ma t h bb {Z}_ p$-module homomor p hi s m $H\to H,h\ma psto ( X, h )$ . \ Th e k ernel of t hi s map is prec i se l y $Z( G )$, therefore $(X,H)\con g \fr ac{H}{ Z( G)} $ . So s ince$Z ( G)\ cap (X,H)=1 $, i t follows that$ Z(G)\ti m es (X,H )$ has th e s amer an kas$H $ , h e nc e i t is open in $ H$ .\ Re call f r o m [@ Bil ly D efini tion 1.6] the de fini t ion of t he *i sola to r* $i _G(N)$ of a c losed, normal s ubgr oup $N$ o f $ G$ , a nd reca l l from [@ Bil ly Prop osition 1.7 ,L e m ma 1.8] that it is a c l o se d, isola ted no r ma ls ubgroupof $G $, a n d that $N$ is open in $i_G( N )$ .\ Let $C =i _G((X, H) )\l eqH$, t h en i t is c lear tha t $Z( G )\cap C=1$ and that $Z(G)\ti m es C $i s is ola ted. There fore , sin ce $ Z (G )\t i mes ( X,H)$ i s o p en in $H$, it follo ws that$Z(G) \times C=H$,and hence$ G = Z(G)\tim es C \ rt i mes\langle X\r angle $, and $G$ has spli t cen tre. *P roof of T h e orem \[B \]. * L et$Z_ 1 ( G) :=\{g\in G:(g , G )\su bs eteq Z( G)\ }$, thi s i s a cl ose dsubgroupof $G$ c on ta in in g $ Z(G)$ and cont ai ned i n $ H$. S u pposefirst tha t$Z _ 1(G )\neq Z ( G) $ . \ Th en c hoos e $ h\ in Z_ 1(G) \ bac kslashZ(G)$, th en$ (h,G )\ su bseteqZ(G)$, so ifwe take $\ps i\ inInn(G) $ to be co njugation by $h$, then

. \[split-centre\] Let $G=H\rtimes\langle_X\rangle$ be_an abelian-by-procyclic group. Then_$G$ has_split_centre if_and_only if $(G,G)\cap_Z(G)=1$. It is clear_that if $G$ has_split centre then_$(G,G)\cap_Z(G)=1$. Conversely, suppose that $(G,G)\cap Z(G)=1$, and consider the $\mathbb{Z}_p$-module homomorphism $H\to H,h\mapsto (X,h)$.\ The_kernel_of this_map_is_precisely $Z(G)$, therefore $(X,H)\cong\frac{H}{Z(G)}$. So_since $Z(G)\cap (X,H)=1$, it follows_that $Z(G)\times_(X,H)$ has the same rank as $H$, hence_it_is open in_$H$.\ Recall from [@Billy Definition 1.6] the definition of the_*isolator* $i_G(N)$ of a closed, normal_subgroup $N$ of_$G$,_and_recall from [@Billy Proposition_1.7, Lemma 1.8] that it is_a closed, isolated normal subgroup of_$G$, and that $N$ is open in_$i_G(N)$.\ Let $C=i_G((X,H))\leq H$, then it is_clear that $Z(G)\cap C=1$ and_that $Z(G)\times_C$ is isolated. Therefore, since $Z(G)\times_(X,H)$ is open_in $H$,_it follows that_$Z(G)\times C=H$, and hence $G=Z(G)\times C\rtimes\langle_X\rangle$, and $G$_has split centre. *Proof of Theorem \[B\].*_Let_$Z_1(G):=\{g\in G:(g,G)\subseteq Z(G)\}$,_this_is_a closed_subgroup of $G$_containing_$Z(G)$ and_contained_in $H$. Suppose first that $Z_1(G)\neq_Z(G)$.\ Then_choose $h\in Z_1(G)\backslash Z(G)$, then $(h,G)\subseteq Z(G)$,_so if we take_$\psi\in_Inn(G)$ to be conjugation_by $h$, then

will never occupy. In @Albers, the pristine region was simply the part of the circular forest between its center and the ring of maximum profit. In general, however, the pristine region must be defined differently. We assume that extractors will extract anywhere where the profit is close to the maximum possible profit. Denote $P_{max}$ as the maximum expected profit, and $(1-\varepsilon)P_{max}$ as the lowest expected profit that the extractor will accept for some tolerance $\varepsilon$. We also assume that the extractors will take the optimal path away from their extraction point to the boundary of the region. The pristine region is determined using the following algorithm. 1. Find the equal profit contour corresponding to $(1 - \varepsilon)P_{max}$. The extractors will only extract within this high-profit region. 2. Pick points inside the high-profit region uniformly at random, and calculate the optimal path from that point to the boundary of the region. The optimal path can be found easily by solving the gradient descent equation $${\frac{\mathrm d\vec x}{\mathrm dt}}=-\nabla C_{B(\vec x_0)}(\vec x),$$ where $\vec x(0)=\vec x_0$ and $C_{B(\vec x_0)}(\vec x)$ is the cost function associated with the benefit value at the extraction point $\vec x_0$. 3. The non-pristine region includes the high-profit region, and all points within some (small) distance of one of the optimal paths found in the previous step. The rest of the region is pristine, and is not traversed by extractors. Metrics for Measuring Patrol Effectiveness ------------------------------------------ To measure the effectiveness of a given patrol strategy $\psi$, we use two metrics, although others could be used. A simple measure, as used in @Albers, is to calculate the proportion of the protected area that is pristine, using, $$\textrm{Pristine Proportion} = \frac{\int_{\Omega} \chi(x,y)\,\mathrm d x\,\mathrm d y}{\int_\Omega\,\mathrm dx\,\mathrm dy},$$ where $\chi(x,y)$ is an indicator function which is 1 in the pristine region and 0 in the non-pristine region. Some parts of the protected region may be more valuable to protect than other regions, and we could instead choose to measure the proportion of the net value protected using the

will never occupy. In @Albers, the pristine region was simply the part of the round forest between its plaza and the ring of maximum net income. In general, however, the pristine area must be defined differently. We assume that extractor will extract anywhere where the profit is near to the maximum possible net income. Denote $ P_{max}$ as the maximum expected profit, and $ (1-\varepsilon)P_{max}$ as the lowest expect profit that the centrifuge will accept for some tolerance $ \varepsilon$. We besides assume that the extractors will take the optimum path away from their extraction point to the boundary of the area. The pristine region is specify using the following algorithm. 1. Find the equal net income contour corresponding to $ (1 - \varepsilon)P_{max}$. The extractors will only extract within this high - profit area. 2. Pick point inside the high - profit region uniformly at random, and calculate the optimal path from that point to the boundary of the region. The optimum way can be found easily by solving the gradient origin equation $ $ { \frac{\mathrm d\vec x}{\mathrm dt}}=-\nabla C_{B(\vec x_0)}(\vec x),$$ where $ \vec x(0)=\vec x_0 $ and $ C_{B(\vec x_0)}(\vec x)$ is the cost function consociate with the benefit value at the extraction point $ \vec x_0$. 3. The non - pristine region includes the eminent - profit region, and all points within some (small) distance of one of the optimal paths rule in the previous step. The rest of the region is pristine, and is not traverse by extractors. prosody for Measuring Patrol Effectiveness ------------------------------------------ To measure the effectiveness of a given patrol strategy $ \psi$, we use two metric unit, although others could be used. A simple measure, as use in @Albers, is to calculate the proportion of the protected area that is pristine, use, $ $ \textrm{Pristine Proportion } = \frac{\int_{\Omega } \chi(x, y)\,\mathrm d x\,\mathrm d y}{\int_\Omega\,\mathrm dx\,\mathrm dy},$$ where $ \chi(x, y)$ is an index function which is 1 in the pristine region and 0 in the non - pristine region. Some parts of the protected area may be more valuable to protect than other regions, and we could instead choose to measure the symmetry of the net value protected using the

wipl never occupy. In @Alberr, the pristine tetion wes simpmy the pxrt of the circular forest bxtwewn itw center and the ring uf maximul profit. In jeneral, however, vge prisbnne rsnion knst be defined cifferentlf. We assume thdt ertractors will extract anywhere wherq the ptovit is close tj tht mwximhm possible profit. Denote $P_{max}$ as fhe maxpmum expected progit, and $(1-\varepsilon)P_{max}$ as hhe powest expected prlfit that tye evrractor will accept fog some toletance $\varepsilon$. We also assume ghat che extractirw wlnl take the optiial path awan from dheir ectraction poinb to vhe voundary of the regioi. The pristine region is deterkiued using the followibg algotithm. 1. Five tfe tquel lrofit coitour correaponding to $(1 - \varepsilon)P_{max}$. Tne vctractors wiml onlr qxtract within this high-profit region. 2. Kick loints inside the high-peofit region uniformli at randoi, and calculate the optimal path from that point do thx couueary ud hhe region. The optimal path can be found easijg ny solving the gvadient descent eqiahipg $${\frac{\mathrm a\vec x}{\matgrm dt}}=-\nabla C_{B(\vec d_0)}(\vec x),$$ rhere $\vec x(0)=\vec x_0$ amd $C_{B(\vec x_0)}(\vec x)$ is the cost function asficiated with the bznefit value at jhe exyraction point $\vec x_0$. 3. Tke non-lristine rehion inclhaes the high-profkt gegimn, and all points within sjme (small) discance of one of thq optimal oaths found in the previlus sjep. Tha rest of hhe region is pristine, and is not traversed bi eftrdctors. Mecrics nor Measuring Pwtrol Effectivgness ------------------------------------------ To mzasure the effecniveness mf a given [atrol strateci $\psi$, we use vwo metriss, aothoygh othdfs could be usrd. A simple measure, as used in @Albers, is fo calculate thz kriportion of thr pfotqcned awaa that is psistkne, ising, $$\textrm{Kxisbind Prpportion} = \frac{\int_{\Omeca} \cgi(x,y)\,\mathrm d x\,\mathtm d y}{\int_\Onega\,\mathwm dx\,\mathrm du},$$ where $\chi(x,y)$ is aj indmcator funcyiog which is 1 in the pristine refion and 0 in the non-priseine region. Some 'arts of the protected region may be morx valuable to protect jhan other regions, anb ee could inxtead choose tm measure the proporrion of the net vslue protected using fhe

will never occupy. In @Albers, the pristine simply part of circular forest between of profit. In general, the pristine region be defined differently. We assume that will extract anywhere where the profit is close to the maximum possible profit. $P_{max}$ as the maximum expected profit, and $(1-\varepsilon)P_{max}$ as the lowest expected profit the will for tolerance $\varepsilon$. We also assume that the extractors will take the optimal path away from their point to the boundary of the region. The region is determined using following algorithm. 1. Find the profit corresponding to - The will only extract this high-profit region. 2. Pick points inside the high-profit region uniformly at random, and calculate the optimal from that the boundary the The path can be by solving the gradient descent equation dt}}=-\nabla C_{B(\vec x_0)}(\vec x),$$ where $\vec x(0)=\vec x_0$ $C_{B(\vec x_0)}(\vec is the cost function associated with benefit value at the extraction point $\vec x_0$. The non-pristine region includes the high-profit region, and all points within some (small) distance of the optimal paths found the previous step. rest the is and is traversed by extractors. Metrics for Measuring Patrol Effectiveness ------------------------------------------ To measure effectiveness of a given patrol strategy $\psi$, we use two others be used. A measure, as used in is calculate the proportion of area is Proportion} \frac{\int_{\Omega} d x\,\mathrm d y}{\int_\Omega\,\mathrm dy},$$ where $\chi(x,y)$ is an function which is 1 0 in the non-pristine region. Some parts of protected region may be more valuable to than other regions, and we could instead choose to measure the proportion the net using the

will never occupy. In @Albers, thE pristine rEgion Was SimPlY the Part Of the circular fORest Between its center and the Ring oF mAXimuM PrOfit. IN generaL, HoWEVer, ThE pRisTiNE rEgion MusT be defiNed differeNtlY. WE assume that eXTrActors will ExtRact anywhere WheRe the pRoFit IS closE to The maXimum pOSsible Profit. DenOtE $p_{max}$ as THe maximUM ExPectEd profit, and $(1-\varepsILoN)p_{max}$ as the lowesT expecTeD PrOFIt tHat The extractOr Will aCCept for SOmE TOLerANce $\varepsilon$. we also assumE ThaT the exTrActORs will Take tHe OPtiMal path away From Their extrAction POint to tHE boundaRy of thE reGioN. The PRiStIne ReGIon IS dEteRMinEd using tHe FoLlowiNg alGORIThm. 1. FInd The eQual pRofit contour cOrrEspoNDinG to $(1 - \vaRepsiLon)P_{MaX}$. The eXtractOrs wiLl Only extract withIn thIs high-proFit ReGioN. 2. PIck poINts insIde The High-proFit regiON unIfORMLy At random, and calculaTe THE oPtimal paTh from THaT pOInt to the BoUndAry oF THe regIon. THE oPtimal paTh can bE FoUnD easily By SolvinG tHe gRadIent dEScenT equatIon $${\frac{\mAthrm D\Vec x}{\mathrm dt}}=-\naBLa C_{B(\vec x_0)}(\vec x),$$ wHErE $\VEc X(0)=\Vec x_0$ And $c_{B(\vec x_0)}(\vec x)$ iS the COst fUnctIOn AssOCiateD with ThE BeNEfit value at the extraCtIon poiNt $\vec X_0$. 3. The non-pristiNe region inCLUDes the hiGh-prOFiT Region, and all poInts wIthin some (sMAll) distaNce of One of the Optimal paTHS found in The PreVioUs sTEP. THe rest of the reGIOn is PrIstine, aNd iS not traVerSed By eXtrAcTors. MetriCs for MeaSuRiNg paTroL EffeCTiveness ------------------------------------------ to MeaSuRe tHe effECtivenEss of A givEn PaTRol StrategY $\PsI$, WE use TwO mEtriCs, aLtHough OtheRS coUld be usEd. A simple MeaSUre, aS uSeD in @AlbeRs, is to calculaTe The proportIoN of The proTECted area That is pristine, using, $$\textRM{PristiNe PRoporTion} = \Frac{\int_{\OmEga} \Chi(x,y)\,\mAthRM d x\,\matHrm d y}{\iNt_\OmeGa\,\MatHRM dx\,\maTHRm Dy},$$ wHeRe $\chi(x,y)$ is aN INdiCator FuNctiOn which Is 1 in the pristine regIOn aNd 0 in the non-priStiNe reGIOn. somE PaRTs oF tHE prOTEcted region may bE more valuaBlE To Protect thaN OthEr Regions, And we coUld inSTead choOse to measUre the proPoRtioN OF thE net value pRotected Using the

will never occupy. In @Al bers, theprist ine re gi on w as s imply the part of t he circular forest bet weenit s cen t er andthe rin g o f max im um pr of i t. In g ene ral, ho wever, the pr is tine regionm us t be defin eddifferently. We assum etha t extr act ors w ill ex t ract a nywhere w he r e thep rofit i s cl oseto the maximum po s si b le profit. Den ote $P _{ m ax } $ as th e maximumex pecte d profit , a n d $(1 - \varepsilon)P _{max}$ ast helowest e xpe c ted pr ofitth a t t he extracto r wi ll accept for s o me tole r ance $\ vareps ilo n$. Wea ls oass um e th a tthe ext ractorswi ll take the o p t imal pa th a way f rom their ext rac tion poi nt to theboun da ry of the r egion . The pristine r egio n is dete rmi ne d u si ng th e follo win g a lgorith m. 1.Fin dt h e e qual profit contou rc o rr espondin g to $ ( 1-\ varepsil on )P_ {max } $ . The ext r ac tors wil l only ex tr act wit hi n this h igh -pr ofitr egio n. 2. Pick p oints inside the hig h -profit regio n u n i fo r mlyatrandom, and cal c ulat e th e o pti m al pa th fr om th a t point to the boun da ry ofthe r egion. The op timal path c a n be fou nd e a si l y by solving t he gr adient des c ent equa tion$${\frac {\mathrmd \ vec x}{\ mat hrm dt }}= - \ na bla C_{B(\vec x _0)} (\ vec x), $$where $ \ve c x (0) =\v ec x_0$ and $C_{B(\ ve cx_ 0) }(\ vec x ) $ is the c ost f unc tiona ssocia ted w ithth eb ene fit val u ea t the e xt ract ion p oint$\ve c x_ 0$. 3. The non -pr i stin ere gion in cludes the hi gh -profit re gi on, and a l l pointswithin some (small) dis t ance of on e oftheoptimal p ath s foun d i n the p reviou s ste p. Th e resto f t here gion is pr i s tin e, an dis n ot trav ersed by extractor s . Metrics for M eas urin g Pa tro l E f fec ti v ene s s -------------- ---------- -- - -- ---------- - -- T o measu re theeffec t iveness of a giv en patrol s trat e g y $ \psi$, weuse twometrics,a lthou g hother s c ould b euse d. Asimple mea sure, as us ed in @A lbers ,is to ca lculate the proportionof the prot ect ed area t hat ispristine, usi ng, $$\tex trm {Pr istin e P r oport ion} =\fr a c{\in t_{\ O mega} \ch i (x ,y) \ , \m athrm d x\, \ m a thr m d y }{\ i nt_\Om ega\ ,\mathrm dx\,\mat h rm dy},$$ wher e $\ c h i(x ,y) $ isan indicator fun cti on w hich is1in the pris tine reg io n and0 in t he non -pristi n e r e gion. Som e p arts of t hepr o tectedre gi o n maybe m or e valu able t o pro t e ct than other re gions , and w e co uld i ns tead ch o oseto measure the propor tion o f th e net valuepr otecte d u si ng the

will_never occupy._In @Albers, the pristine_region was_simply_the part_of_the circular forest_between its center_and the ring of_maximum profit. In_general,_however, the pristine region must be defined differently. We assume that extractors will extract_anywhere_where the_profit_is_close to the maximum possible_profit. Denote $P_{max}$ as the_maximum expected_profit, and $(1-\varepsilon)P_{max}$ as the lowest expected profit_that_the extractor will_accept for some tolerance $\varepsilon$. We also assume that_the extractors will take the optimal_path away from_their_extraction_point to the boundary_of the region. The pristine region is_determined using the following algorithm. 1. _Find the equal profit contour corresponding to_$(1 - \varepsilon)P_{max}$. The extractors will_only extract within this high-profit_region. 2. _Pick points inside the high-profit_region uniformly at_random, and_calculate the optimal_path from that point to the_boundary of the_region. The optimal path can be_found_easily by solving_the_gradient_descent equation_$${\frac{\mathrm d\vec x}{\mathrm_dt}}=-\nabla_C_{B(\vec x_0)}(\vec_x),$$_where $\vec x(0)=\vec x_0$ and $C_{B(\vec_x_0)}(\vec_x)$ is the cost function associated with_the benefit value at_the_extraction point $\vec x_0$. 3._ The non-pristine region includes_the high-profit region, and all points_within some_(small) distance_of one of the optimal paths found in the previous step._The rest of the region is_pristine, and is not_traversed by_extractors. Metrics_for Measuring Patrol_Effectiveness ------------------------------------------ To_measure the_effectiveness of a given patrol strategy $\psi$,_we use_two metrics, although others could be_used. A simple measure,_as_used in @Albers, is to calculate_the proportion of the protected area_that is pristine, using, $$\textrm{Pristine_Proportion}_=_\frac{\int_{\Omega} \chi(x,y)\,\mathrm d x\,\mathrm d_y}{\int_\Omega\,\mathrm dx\,\mathrm dy},$$ where $\chi(x,y)$ is_an indicator function_which is 1 in the pristine region_and_0 in the non-pristine region. Some parts_of_the protected region may be more_valuable_to_protect than other regions, and_we could instead choose to measure_the proportion of the net value protected using the

use one of the most common methods of combining two types of features, which concatenates them together and applies a MLP. As we have mentioned, the combination result, as the new spatial feature, is actually more comprehensive and contains both spatial and structural information. Therefore, in the pipeline of our network, we concatenate all the combination results for generating global feature. #### Aggregation of Structural Feature With the input structural feature and neighbor index, we aggregate the feature of a face with features of its connected faces. Several aggregation methods are listed and discussed as follows: - **Average pooling:** The average pooling may be the most common aggregation method, which simply calculates the average value of features in each channel. However, this method sometimes weakens the strongly-activated areas of the original features and reduce the distinction of them. - **Max pooling:** Another pooling method is the max pooling, which calculates the max value of features in each channel. Max pooling is widely used in 2D and 3D deep learning frameworks for its advantage of maintaining the strong activation of neurons. - **Concatenation:** We define the concatenation aggregation, which concatenates the feature of face with feature of each neighbor respectively, passes these pairs through a shared MLP and applies a max pooling to the results. This method both keeps the original activation and leaves space for the network to combine neighboring features. We finally use the concatenation method in this paper. After aggregation, another MLP is applied to further fusion the neighboring features and output new structural feature. Implementation Details ---------------------- Now we present the details of implementing MeshNet, illustrated in Fig 2, including the settings of hyper-parameters and some details of the overall architecture. The spatial descriptor contains fully-connected layers (64, 64) and output a initial spatial feature with length of 64. Parameters inside parentheses indicate the dimensions of layers except the input layer. In the face rotate convolution, we set $K_1=32$ and $K_2 = 64$, and correspondingly, the functions $f(\cdot, \cdot)$ and $g(\cdot)$ are implemented by fully-connected layers (32, 32) and (64, 64). In the face kernel correlation, we set $M = 64$ (64 kernels) and $\sigma = 0.2$. We parameterize the mesh convolution block with a four-tuple $(in_1, in_2, out_1,

use one of the most common methods of compound two type of features, which concatenates them together and apply a MLP. As we have mentioned, the combination result, as the modern spatial feature of speech, is actually more comprehensive and contains both spatial and structural information. Therefore, in the pipeline of our net, we concatenate all the combination results for generating global feature. # # # # collection of Structural Feature With the input structural feature and neighbor index, we aggregate the feature of a expression with features of its connected faces. respective aggregation methods are number and discussed as come: - * * Average pooling :* * The average pooling may be the most common aggregation method acting, which simply calculates the average value of features in each channel. However, this method sometimes de-escalate the powerfully - activated areas of the original features and reduce the distinction of them. - * * Max pooling :* * Another pooling method is the max pooling, which calculates the soap value of features in each distribution channel. Max pooling is widely used in 2D and 3D deep memorize frameworks for its advantage of maintaining the strong energizing of neurons. - * * chain :* * We define the concatenation aggregation, which concatenate the feature of face with feature of each neighbor respectively, passes these pairs through a shared MLP and applies a max pooling to the results. This method both keeps the original activation and bequeath outer space for the network to combine neighboring features. We finally use the chain method in this paper. After aggregation, another MLP is applied to further fusion the neighboring features and output new structural feature. Implementation Details ---------------------- immediately we present the detail of implementing MeshNet, exemplify in Fig 2, including the setting of hyper - parameters and some contingent of the overall architecture. The spatial form contains fully - connected layer (64, 64) and output a initial spatial feature with distance of 64. Parameters inside parentheses indicate the dimensions of layers except the input layer. In the face rotate convolution, we put $ K_1=32 $ and $ K_2 = 64 $, and correspondingly, the functions $ f(\cdot, \cdot)$ and $ g(\cdot)$ are implemented by in full - connected layers (32, 32) and (64, 64). In the font kernel correlation, we set $ M = 64 $ (64 kernel) and $ \sigma = 0.2$. We parameterize the mesh convolution block with a four - tuple $ (in_1, in_2, out_1,

usf one of the most common methods of comyuning vwo typss of fextures, which concatenates thxm tigethtg and applies a MLP. Ar we have mentionwd, tie combination rxault, as the nsa spctmal feature, is sctually mmre comprehenshvd cnd contains both spatial and structtral ingogmation. Therefjre, pn the iikeline of our network, we concatenzte all the combinatoon results for generating glohal feature. #### Aggregwtion of Stthctteal Feature Wkth the inkuc structuram feature and neighbor index, we aggrzgate the fgcrurf of a face xith fvatures of its connecded facrs. Several aggvegatmon nethods are listed anv discussed as follors: - **Averdgz pooling:** The average piolinc maf be rhe moat ckmmon wggcegation mefhod, which wimply calculates tne qverage value of fewttres in each channel. However, this methog skmetimes weakens the steongly-activated areas of the owiginal features and reduce the distinction of tham. - **Jxx koiling:** Qnlther pooling method is the max pooling, which dakcllates the max vajue of featirfs yn each channgl. Max 'koming is widely usef in 2D wnd 3D deep leawninb frameworks for its advantqge of maintcinung the strong actnvation of nzurons. - **Comcatenation:** We define tke condatenation wggregatikv, which concatenxtex dhe feature of face with fqature of eack neighbur rgspectidely, passed these pairs through a dhareb MLP and applifs a max pooling to the results. This method bptv kveps the jrigikal activation wnd leaves spaee for tke netdork to cojbine nxighboring fqatures. We findply use the roncatenaeion metyod in gfis paper. Aftet aggreganijn, abother MLP is applled tu further fusion tht beighboring festufes ajd otdput new strgctufal geatufe. Implementctkon Cetails ---------------------- Now we presend ths details of implekekting MesyNet, illtstrated in Fog 2, including the dettiigs of hypet-patameters and some details of ths overall arghitecture. The spabial descriptox contains fully-connected layers (64, 64) and ontput a initial spatial feature with lengtk pf 64. Parametxrs infide parettheses indicate the dimensions of laners except the input mayer. Hn thf face rotate convolution, we set $K_1=32$ and $K_2 = 64$, and correspondingly, the functiins $f(\cvoe, \cdot)$ and $f(\cdoy)$ are ii'lementqd bb hully-connected lajers (32, 32) and (64, 64). In the face kernel rorrelatiot, ce set $M = 64$ (64 kernels) and $\sigma = 0.2$. De parameteriee the mesh convolution blkck with a four-tuple $(in_1, in_2, out_1,

use one of the most common methods two of features, concatenates them together we mentioned, the combination as the new feature, is actually more comprehensive and both spatial and structural information. Therefore, in the pipeline of our network, we all the combination results for generating global feature. #### Aggregation of Structural Feature the structural and index, we aggregate the feature of a face with features of its connected faces. Several aggregation are listed and discussed as follows: - **Average The average pooling may the most common aggregation method, simply the average of in channel. However, this sometimes weakens the strongly-activated areas of the original features and reduce the distinction of them. - **Max Another pooling the max which the value of features channel. Max pooling is widely used 3D deep learning frameworks for its advantage of the strong of neurons. - **Concatenation:** We define concatenation aggregation, which concatenates the feature of face feature of each neighbor respectively, passes these pairs through a shared MLP and applies a to the results. This both keeps the activation leaves for network to neighboring features. We finally use the concatenation method in this paper. aggregation, another MLP is applied to further fusion the neighboring output structural feature. Implementation ---------------------- Now we present details implementing MeshNet, illustrated in including settings some of overall architecture. The spatial contains fully-connected layers (64, 64) output a initial spatial Parameters inside parentheses indicate the dimensions of layers the input layer. In the face rotate we set $K_1=32$ and $K_2 = 64$, and correspondingly, the functions $f(\cdot, and $g(\cdot)$ by fully-connected layers (32, 32) and (64, 64). the face kernel correlation, set $M = 64$ (64 kernels) and $\sigma = We the mesh block with a $(in_1, in_2, out_1,

use one of the most common methOds of combiNing tWo tYpeS oF feaTureS, which concatenATes tHem together and applies a mLP. As We HAve mENtIoned, The combINaTIOn rEsUlT, as ThE NeW spatIal Feature, Is actually MorE cOmprehensive ANd Contains boTh sPatial and strUctUral inFoRmaTIon. ThEreFore, iN the piPEline oF our netwoRk, WE concaTEnate alL THe CombInation results for GEnERating global feAture. #### AGgREgATIon Of STructural FEaTure WITh the inPUt STRUctURal feature and Neighbor indEX, we AggregAtE thE FeaturE of a fAcE WitH features of Its cOnnected fAces. SeVEral aggREgation MethodS arE liSted ANd DiScuSsED as FOlLowS: - **aveRage poolInG:** THe aveRage POOLIng mAy bE the Most cOmmon aggregatIon MethOD, whIch siMply cAlcuLaTes thE averaGe valUe Of features in eacH chaNnel. HowevEr, tHiS meThOd somETimes wEakEns The stroNgly-actIVatEd AREAs Of the original featuReS ANd Reduce thE distiNCtIoN Of them. - **MaX pOolIng:** ANOTher pOoliNG mEthod is tHe max pOOlInG, which cAlCulateS tHe mAx vAlue oF FeatUres in Each chanNel. MaX Pooling is widelY Used in 2D and 3D deEP lEARnINg frAmeWorks for its AdvaNTage Of maINtAinINg the StronG aCTiVAtion of neurons. - **ConcaTeNation:** we defIne the concateNation aggrEGATion, whicH conCAtENates the featurE of faCe with featURe of each NeighBor respeCtively, paSSEs these pAirS thRouGh a SHArEd MLP and appliES A max PoOling to The Results. thiS meThoD boTh Keeps the oRiginal aCtIvAtIoN anD leavES space foR tHe nEtWorK to coMBine neIghboRing FeAtURes. we finalLY uSE The cOnCaTenaTioN mEthod In thIS paPer. AfteR aggregatIon, ANothEr mLp is applIed to further fUsIon the neigHbOriNg featUREs and outPut new structural feature. iMplemenTatIon DeTailS ---------------------- Now we preSenT the deTaiLS of impLementIng MeShnet, ILLustrATEd In FIg 2, Including tHE SetTings Of HypeR-parameTers and some details OF thE overall archiTecTure. tHE sPatIAl DEscRiPTor CONtains fully-connEcted layerS (64, 64) aND oUtput a initIAl sPaTial feaTure witH lengTH of 64. ParaMeters insIde parentHeSes iNDIcaTe the dimenSions of lAyers excePT the iNPuT layeR. In The facE rOtaTe conVolutiON, we Set $K_1=32$ aNd $K_2 = 64$, and CoRrespoNdingLy, The functIons $f(\cdot, \cdot)$ and $g(\cdot)$ arE impleMenteD by Fully-connEctED laYers (32, 32) and (64, 64). In The fAce kernel cOrrElaTion, wE seT $m = 64$ (64 kernEls) aND $\sIgmA = 0.2$. we parAmetERize the meSH cOnvOLUtIon block witH A FOur-Tuple $(In_1, iN_2, Out_1,

use one of the most commo n methodsof co mbi nin gtwotype s of features, whic h concatenates them to gethe ra nd a p pl ies a MLP. A s w e hav eme nti on e d, thecom binatio n result,asth e new spatia l f eature, is ac tually morecom prehen si vea nd co nta ins b oth sp a tial a nd struct ur a l info r mation. T he refo re, in the pipeli n eo f our network, we co nc a te n a teall the combi na tionr esultsf or g e ner a ting global f eature. ## # # A ggrega ti ono f Stru ctura lF eat ure With t he i nput stru ctural feature and nei ghborind ex, wea gg re gat et hef ea tur e of a facewi th feat ures o f itscon nect ed fa ces. Severalagg rega t ion meth ods a re l is ted a nd dis cusse das follows: - ** Average p ool in g:* *The a v eragepoo lin g may b e the m o stco m m o naggregation method ,w h ic h simply calcu l at es the aver ag e v alue o f fea ture s i n each c hannel . H ow ever, t hi s meth od so met imesw eake ns the strongl y-act i vated areas of the originalf ea t u re s and re duce the di stin c tion oft he m.- * *Maxpo o li n g:** Another poolin gmethod is t he max poolin g, which c a l c ulates t he m a xv alue of featur es in each chan n el. Maxpooli ng is wi dely used i n 2D and 3D de eplea r n in g frameworksf o r it sadvanta geof main tai nin g t hest rong acti vation o fne ur on s. - * *Concate na tio n: **We de f ine th e con cate na ti o n a ggregat i on , whic hco ncat ena te s the fea t ure of fac e with fe atu r e of e ac h neigh bor respectiv el y, passesth ese pairs t hrough a shared MLP and applies a max p ool ing t o th e results . T his me tho d bothkeepsthe o ri gin a l acti v a ti onan d leaves s p a cefor t he net work to combine neighbori n g f eatures. Wefin ally u se th e c o nca te n ati o n method in this paper. Af te r a ggregation , an ot her MLP is app liedt o furth er fusion the neig hb orin g fea tures andoutput n ew struct u ral f e at ure. Im plemen ta tio n Det ails - - --- ----- ------ -- ---- Now w epresentthe details of implemen ting M eshNe t,illustrat edi n F ig 2, inc ludi ng the set tin gsof hy per - param eter s a nds ome d etai l s of theo ve ral l ar chitecture. T hespati ald escrip torcontains fully-co n nected layers(64, 6 4)and outp ut a initial spa tia lf e ature wi th length of64. Para me t ers i nsideparent heses i n d ic a te the dim ens ions of l aye rs exceptth ei nput l ayer .In the facer otat e convolution, weset $ K _ 1=32$ and $K_2 = 64$, a n d co rrespondin gly, the fu nction s $f (\cdo t, \cdo t) $ and$g( \c dot)$ arei mplemente d byfully-c on nect edlayers (32 , 32) a nd ( 64 , 6 4). In th e fa c eke r nel cor relat io n, w e set $M= 64$ (64 ke r nels) a nd $\ s i gma =0 .2 $ . We param ete rizet h e mesh con v olut i on block witha four -tuple$ (in _1 , in_2, ou t _ 1,

use_one of_the most common methods_of combining_two_types of_features,_which concatenates them_together and applies_a MLP. As we_have mentioned, the_combination_result, as the new spatial feature, is actually more comprehensive and contains both spatial_and_structural information._Therefore,_in_the pipeline of our network,_we concatenate all the combination_results for_generating global feature. #### Aggregation of Structural Feature With the_input_structural feature and_neighbor index, we aggregate the feature of a face_with features of its connected faces._Several aggregation methods_are_listed_and discussed as follows: -_ **Average pooling:** The average_pooling may be the most common_aggregation method, which simply calculates the average_value of features in each channel._However, this method sometimes weakens_the strongly-activated_areas of the original features_and reduce the_distinction of_them. - _**Max pooling:** Another pooling method is_the max pooling,_which calculates the max value of_features_in each channel._Max_pooling_is widely_used in 2D_and_3D deep_learning_frameworks for its advantage of maintaining_the_strong activation of neurons. - **Concatenation:**_We define the concatenation_aggregation,_which concatenates the feature_of face with feature of_each neighbor respectively, passes these pairs_through a_shared MLP_and applies a max pooling to the results. This method both_keeps the original activation and leaves_space for the network_to combine_neighboring_features. We finally use_the_concatenation method_in this paper. After aggregation, another MLP_is applied_to further fusion the neighboring features_and output new structural_feature. Implementation_Details ---------------------- Now we present the details of_implementing MeshNet, illustrated in Fig 2,_including the settings of hyper-parameters_and_some_details of the overall architecture. The_spatial descriptor contains fully-connected layers (64,_64) and output_a initial spatial feature with length of_64._Parameters inside parentheses indicate the dimensions_of_layers except the input layer. In_the_face_rotate convolution, we set $K_1=32$_and $K_2 = 64$, and correspondingly,_the functions $f(\cdot, \cdot)$ and $g(\cdot)$ are implemented by_fully-connected layers (32,_32) and (64, 64). In_the_face_kernel correlation, we set $M = 64$ (64 kernels) and_$\sigma =_0.2$. We parameterize the_mesh convolution block with a four-tuple $(in_1, in_2, out_1,

, Tom M. Herbst, Stanimir Metchev' title: Ruling out Stellar Companions and Resolving the Innermost Regions of Transitional Disks with the Keck Interferometer --- Introduction ============ Circumstellar disks are a natural outcome of the star-formation process: when a molecular cloud core collapses, it gives rise to a central star surrounded by a rotating circumstellar disk, which transports material towards the star. Over time, the disk material dissipates through [ processes such as]{} accretion onto the central star, [ disk winds]{} and the formation of planets. At an age of $\sim5~{\rm Myr}$, about 90% of disks have already dispersed, and within $10~{\rm Myr}$ of their formation, almost all pre-main-sequence stars are diskless [e.g. @2006ApJ...638..897S]. While it is now believed that such disks commonly give rise to planetary systems, the details of this process remain unclear. Theory predicts that disks evolve from the inside out: dust grain growth is expected to occur faster in the inner disk than in the outer disk, higher densities favor planet formation in the inner disk [@2002Icar..156..291B], and photoevaporation by the central star will cause the inner disk to dissipate first [@2001MNRAS.328..485C]. [ Possible observational support for inside out disk evolution has been found in a small number of so-called transitional disks.]{} These systems show a strong mid-infrared excess ($\gtrsim\,8\,\mu {\rm m}$) revealing the presence of dust [ but significantly reduced or no shorter wavelength infrared excess compared to typical classical T Tauri disks, indicating a depletion of optically thick inner dust out to a radius of a few AU]{} [@1990AJ.....99.1187S; @1997ApJ...489L.173M; @2001AJ....121.1003S; @2005ApJ...621..461D; @2005ApJ...630L.185C; @2006ApJ...636..932M; @2006ApJS..165..568F; @2006AJ....131.1574L; @2006ApJ...643.1003M; @2007ApJ...664L.111E_disk; @2007ApJ...664L.107B]. Therefore, these disks might be in the process of dispersing and this has often been assumed to be due to the

, Tom   M.   Herbst, Stanimir   Metchev' title: Ruling out Stellar Companions and Resolving the Innermost Regions of Transitional Disks with the Keck Interferometer --- initiation = = = = = = = = = = = = Circumstellar phonograph record are a natural outcome of the star - geological formation process: when a molecular cloud kernel collapses, it gives ascent to a central star wall by a rotating circumstellar disk, which enchant material towards the star. Over time, the disk material dissipates through [ procedure such as ] { } accretion onto the central star, [ disk wind ] { } and the formation of planets. At an age of $ \sim5~{\rm Myr}$, about 90% of disk have already dispersed, and within $ 10~{\rm Myr}$ of their formation, almost all pre - main - sequence asterisk are diskless [ e.g. @2006ApJ... 638.. 897S ]. While it is now believed that such disks commonly give rise to planetary systems, the details of this process remain unclear. Theory predicts that disks evolve from the inside out: debris texture growth is expected to occur quicker in the inside disk than in the outer disk, higher densities favor planet formation in the inside disk [ @2002Icar.. 156.. 291B ], and photoevaporation by the central star will cause the inner disk to dissipate first [ @2001MNRAS.328.. 485C ]. [ Possible observational support for inside out disk development has been found in a little numeral of so - called transitional disks. ] { } These systems usher a strong mid - infrared excess ($ \gtrsim\,8\,\mu { \rm m}$) revealing the bearing of dust [ but importantly reduced or no shorter wavelength infrared excess compared to typical authoritative T Tauri disks, indicating a depletion of optically thick inner dust out to a spoke of a few AU ] { } [ @1990AJ..... 99.1187S; @1997ApJ... 489L.173 M; @2001AJ.... 121.1003S; @2005ApJ... 621.. 461D; @2005ApJ... 630L.185C; @2006ApJ... 636.. 932 M; @2006ApJS.. 165.. 568F; @2006AJ.... 131.1574L; @2006ApJ... 643.1003 M; @2007ApJ... 664L.111E_disk; @2007ApJ... 664L.107B ]. Therefore, these disks might be in the process of dispersing and this has much been assumed to be due to the

, Tol M. Herbst, Stanimir Metchev' uitle: Ruling out Stellar Compahions ana Resolving the Innermost Rejionw of Ugansitional Disks witf the Kecn Interfwromtter --- Introduction ============ Rjrcumstcjlar fiskv are a natural outcome ox the star-formdtkou process: when a molecular cloud corq collalsfs, it gives rife tp a csntral star surrounded by a rotatihg circlmstellar disk, whoch transports material toaardd the star. Over tile, the disk matqeial dissipages through [ processes such as]{} accretion onto the cengral xtar, [ disk wundd]{} and the focmatiog of planets. At an ace of $\som5~{\rm Myr}$, about 90% oh diwks have already dispxrsed, and within $10~{\rm Iyr}$ of thaix formation, almost alo pre-mahn-sexuenzw sgara ere disklfss [e.g. @2006ApJ...638..897S]. Wgile it is bow believed that sicr disks commonmy givq wise to planetary systems, the details ox tgis process remain unclwar. Theory predicts tjat disks evolve from the inside out: dust grain growth is axpecvea ti jzxug faster in the inner disk than in the outer qjsl, migher densities favor planrt fptmation in the inner diak [@2002Icar..156..291B], and photofvaporajion bt the cenural xtar will cause the inner dusk to dissiiate first [@2001MNRAS.328..485C]. [ Possnble observacional suppprt for inside out disk evomution has heen founs in a small numbdr pf so-called transitional dishs.]{} These wystzms show a sjrong myd-infrared excess ($\gtrsim\,8\,\mu {\rm m}$) regealiug tha presence of dust [ but significantly rednred or no shottes wdvelengtk infrsred excess cjmpared to typncal clafsicau T Tauri sisks, iidicating a qepletion of mktically thicn inner qust out to a rxaius of a few SU]{} [@1990AJ.....99.1187S; @1997ApL...489J.173M; @2001AH....121.1003S; @2005ApJ...621..461D; @2005ApJ...630L.185C; @2006ApJ...636..932M; @2006ApGA..165..568F; @2006AJ....131.1574L; @2006ApJ...643.1003M; @2007ApJ...664O.111E_dusk; @2007ApJ...664L.107B]. Theregord, trede dyvks might be in ghe lrocers of dispevsivg amd this has often bean aasumed to be due tp bhe

, Tom M. Herbst, Stanimir Metchev' title: Stellar and Resolving Innermost Regions of Interferometer Introduction ============ Circumstellar are a natural of the star-formation process: when a cloud core collapses, it gives rise to a central star surrounded by a circumstellar disk, which transports material towards the star. Over time, the disk material through processes as]{} onto the central star, [ disk winds]{} and the formation of planets. At an age of Myr}$, about 90% of disks have already dispersed, within $10~{\rm Myr}$ of formation, almost all pre-main-sequence stars diskless @2006ApJ...638..897S]. While is believed such disks commonly rise to planetary systems, the details of this process remain unclear. Theory predicts that disks evolve from inside out: growth is to faster the inner disk the outer disk, higher densities favor the inner disk [@2002Icar..156..291B], and photoevaporation by the star will the inner disk to dissipate first [ Possible observational support for inside out disk has been found in a small number of so-called transitional disks.]{} These systems show a excess ($\gtrsim\,8\,\mu {\rm m}$) the presence of [ significantly or shorter wavelength excess compared to typical classical T Tauri disks, indicating a depletion optically thick inner dust out to a radius of a [@1990AJ.....99.1187S; @2001AJ....121.1003S; @2005ApJ...621..461D; @2005ApJ...630L.185C; @2006ApJS..165..568F; @2006AJ....131.1574L; @2006ApJ...643.1003M; @2007ApJ...664L.111E_disk; Therefore, disks might be in of and been to due to the

, Tom M. Herbst, Stanimir Metchev' Title: RulinG out STelLar coMpanIons And Resolving thE inneRmost Regions of TransitiOnal DIsKS witH ThE Keck interfeROmETEr --- INtRoDucTiON ============ CIrcumSteLlar disKs are a natuRal OuTcome of the stAR-fOrmation prOceSs: when a molecUlaR cloud CoRe cOLlapsEs, iT giveS rise tO A centrAl star surRoUNded by A RotatinG CIrCumsTellar disk, which trANsPOrts material toWards tHe STaR. oVer TimE, the disk maTeRial dISsipateS ThROUGh [ pROcesses such as]{} Accretion onTO thE centrAl StaR, [ Disk wiNds]{} anD tHE foRmation of plAnetS. At an age oF $\sim5~{\rm mYr}$, about 90% OF disks hAve alrEadY diSperSEd, AnD wiThIN $10~{\rm mYr}$ Of tHEir FormatioN, aLmOst alL pre-MAIN-SequEncE staRs are Diskless [e.g. @2006ApJ...638..897s]. WhIle iT Is nOw belIeved That SuCh disKs commOnly gIvE rise to planetarY sysTems, the deTaiLs Of tHiS procESs remaIn uNclEar. TheoRy prediCTs tHaT DISkS evolve from the insiDe OUT: dUst grain Growth IS eXpECted to ocCuR faSter IN The inNer dISk Than in thE outer DIsK, hIgher deNsIties fAvOr pLanEt forMAtioN in the Inner disK [@2002Icar..156..291b], And photoevaporATion by the centRAl STAr WIll cAusE the inner diSk to DIssiPate FIrSt [@2001MnrAS.328..485C]. [ POssibLe OBsERvational support for InSide ouT disk Evolution has bEen found in A SMAll numbeR of sO-CaLLed transitionaL diskS.]{} These systEMs show a sTrong Mid-infraRed excess ($\GTRsim\,8\,\mu {\rm M}$) reVeaLinG thE PReSence of dust [ buT SIgniFiCantly rEduCed or no ShoRteR waVelEnGth infrarEd excess CoMpArEd To tYpicaL ClassicaL T tauRi DisKs, indICating A deplEtioN oF oPTicAlly thiCK iNNEr duSt OuT to a RadIuS of a fEw AU]{} [@1990aj.....99.1187S; @1997APJ...489L.173M; @2001AJ....121.1003S; @2005apJ...621..461D; @2005ApJ...630L.185C; @2006apJ...636..932m; @2006apJS..165..568f; @2006Aj....131.1574L; @2006apJ...643.1003M; @2007ApJ...664l.111E_disk; @2007ApJ...664L.107B]. ThErEfore, these DiSks Might bE IN the procEss of dispersing and this hAS often bEen AssumEd to Be due to thE

, Tom M. Herbst, StanimirMetchev' t itle: Ru lin goutStel lar Companions andResolving the Innermos t Reg io n s of Tr ansit ional D i sk s wit hth e K ec k I nterf ero meter --- Intro duc ti on ========= = == Circumst ell ar disks are anatura lout c ome o f t he st ar-for m ationprocess:wh e n a mo l ecularc l ou d co re collapses, itg iv e s rise to a ce ntralst a rs u rro und ed by a ro ta tingc ircumst e ll a r dis k , which trans ports mater i altoward sthe star.Overti m e,the disk ma teri al dissip ates t h rough [ process es suc h a s]{ } ac c re ti onon t o t h ecen t ral star, [ d is k win ds]{ } a n d th e f orma tionof planets. A t a n ag e of $\si m5~{\ rm M yr }$, a bout 9 0% of d isks have alrea dy d ispersed, an dwit hi n $10 ~ {\rm M yr} $ o f their format i on, a l m o st all pre-main-sequ en c e s tars are diskl e ss [ e .g. @200 6A pJ. ..63 8 . .897S ]. W h il e it isnow be l ie ve d thatsu ch dis ks co mmo nly g i ve r ise to planeta ry sy s tems, the deta i ls of this pr o ce s s r e main un clear. Theo ry p r edic ts t h at di s ks ev olvefr o mt he inside out: dust g rain g rowth is expectedto occur f a s t er in th e in n er disk than in t he ou ter disk,h igher de nsiti es favor planet f o r mation i n t heinn erd i sk [@2002Icar.. 1 5 6..2 91 B], and ph otoevap ora tio n b y t he centralstar wil lca us ethe inne r disk to d iss ip ate firs t [@200 1MNRA S.32 8. .4 8 5C] . [ Po s si b l e ob se rv atio nal s uppor t fo r in side ou t disk ev olu t ionha sbeen fo und in a smal lnumber ofso -ca lled t r a nsitiona l disks.]{} These syste m s showa s trong mid -infrared ex cess ( $\g t rsim\, 8\,\mu {\rm m }$) r eveal i n gthe p resence of d ust [ bu tsign ificant ly reduced or no s h ort er wavelength in frar e d e xce s sc omp ar e d t o typical classic al T Tauri d i sk s, indicat i ngadepleti on of o ptica l ly thic k inner d ust out t oa ra d i usof a few A U]{} [@1 990AJ.... . 99.11 8 7S ; @19 97A pJ...4 89 L.1 73M;@2001A J ... .121. 1003S; @ 2005Ap J...6 21 ..461D;@2005ApJ...630L.185C; @ 2006Ap J...6 36. .932M; @2 006 A pJS ..165..56 8F;@2006AJ... .13 1.1 574L; @2 0 06ApJ ...6 4 3. 100 3 M; @2 007A p J...664L. 1 11 E_d i s k; @2007ApJ.. . 6 6 4L. 107B] . T h erefor e, t hese disks mightb e in the proce ss o f dis per s ingan d this has oft enbe e n assumed t o be due to the

, Tom M. Herbst,_Stanimir Metchev' title: Ruling_out Stellar Companions and_Resolving the_Innermost_Regions of_Transitional_Disks with the_Keck Interferometer --- Introduction ============ Circumstellar_disks are a natural_outcome of the_star-formation_process: when a molecular cloud core collapses, it gives rise to a central star_surrounded_by a_rotating_circumstellar_disk, which transports material towards_the star. Over time, the_disk material_dissipates through [ processes such as]{} accretion onto_the_central star, [_disk winds]{} and the formation of planets. At an_age of $\sim5~{\rm Myr}$, about 90%_of disks have_already_dispersed,_and within $10~{\rm Myr}$_of their formation, almost all pre-main-sequence_stars are diskless [e.g. @2006ApJ...638..897S]. While_it is now believed that such disks_commonly give rise to planetary systems,_the details of this process_remain unclear._Theory predicts that disks evolve_from the inside_out: dust_grain growth is_expected to occur faster in the_inner disk than_in the outer disk, higher densities_favor_planet formation in_the_inner_disk [@2002Icar..156..291B],_and photoevaporation by_the_central star_will_cause the inner disk to dissipate_first_[@2001MNRAS.328..485C]. [ Possible observational support for inside out_disk evolution has been_found_in a small number_of so-called transitional disks.]{} These_systems show a strong mid-infrared excess_($\gtrsim\,8\,\mu {\rm_m}$) revealing_the presence of dust [ but significantly reduced or no shorter_wavelength infrared excess compared to typical_classical T Tauri disks,_indicating a_depletion_of optically thick_inner_dust out_to a radius of a few AU]{}_[@1990AJ.....99.1187S; @1997ApJ...489L.173M;_@2001AJ....121.1003S; @2005ApJ...621..461D; @2005ApJ...630L.185C; @2006ApJ...636..932M; @2006ApJS..165..568F; @2006AJ....131.1574L;_@2006ApJ...643.1003M; @2007ApJ...664L.111E_disk; @2007ApJ...664L.107B]. Therefore,_these_disks might be in the process_of dispersing and this has often_been assumed to be due_to_the

partial^2_{qq}\mathcal F^{-1}_p(f)) + \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A} \partial^2_{qp}\mathcal F^{-1}_p(f)) + \\ &&\Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A} \partial^2_{pq}\mathcal F^{-1}_p(f)) +\Lambda^{21}\Lambda^{21}\partial^2_{qq}\tilde{A} \partial^2_{pp}\mathcal F^{-1}_p(f))=a\alpha e^{-q}\partial^2_{pp}\mathcal F^{-1}_p(f)=\\ &&P^3(\tilde A,\mathcal F^{-1}_p(f))=(-1)^6a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)=a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)\end{aligned}$$ and $P^k(\tilde A,\mathcal F^{-1}_p(f))=a\alpha e^{-q}\partial^k_{p\dots p}\mathcal F^{-1}_p(f)\quad\forall k \geq 4,$ - If $\tilde A=b\beta e^q+dp.$ $$P^k(\tilde A,\mathcal F^{-1}_p(f))=(-1)^k b\beta e^q\partial^k_{p\dots p}\mathcal F^{-1}_p(f)$$with $\quad\forall k \geq 2 $ - If $ \tilde{A}=a\alpha e^{-q}+b\beta e^q+dp,$ $$\begin{aligned} &&P^2(\tilde{A},\mathcal F^{-1}_p(f)) = \Lambda^{12}\Lambda^{12}\partial^2_{pp}\tilde{A} \partial^2_{qq}\mathcal F^{-1}_p(f)) + \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A} \partial^2_{qp}\mathcal F^{-1}_p(f)) + \\ &&\Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A} \partial^2_{pq}\mathcal F^{-1

partial^2_{qq}\mathcal F^{-1}_p(f) ) + \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A } \partial^2_{qp}\mathcal F^{-1}_p(f) ) + \\ & & \Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A } \partial^2_{pq}\mathcal F^{-1}_p(f) ) + \Lambda^{21}\Lambda^{21}\partial^2_{qq}\tilde{A } \partial^2_{pp}\mathcal F^{-1}_p(f))=a\alpha e^{-q}\partial^2_{pp}\mathcal F^{-1}_p(f)=\\ & & P^3(\tilde A,\mathcal F^{-1}_p(f))=(-1)^6a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)=a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)\end{aligned}$$ and $ P^k(\tilde A,\mathcal F^{-1}_p(f))=a\alpha e^{-q}\partial^k_{p\dots p}\mathcal F^{-1}_p(f)\quad\forall k \geq 4,$ - If $ \tilde A = b\beta e^q+dp.$ $ $ P^k(\tilde A,\mathcal F^{-1}_p(f))=(-1)^k b\beta e^q\partial^k_{p\dots p}\mathcal F^{-1}_p(f)$$with $ \quad\forall k \geq 2 $ - If $ \tilde{A}=a\alpha e^{-q}+b\beta e^q+dp,$ $ $ \begin{aligned } & & P^2(\tilde{A},\mathcal F^{-1}_p(f) ) = \Lambda^{12}\Lambda^{12}\partial^2_{pp}\tilde{A } \partial^2_{qq}\mathcal F^{-1}_p(f) ) + \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A } \partial^2_{qp}\mathcal F^{-1}_p(f) ) + \\ & & \Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A } \partial^2_{pq}\mathcal F^{-1

parhial^2_{qq}\mathcal F^{-1}_p(f)) + \Lamnda^{12}\Lambda^{21}\partial^2_{kq}\rilde{A} \partizl^2_{qp}\mathzal F^{-1}_p(f)) + \\ &&\Lambda^{21}\Lambda^{12}\pactiao^2_{qp}\tiode{A} \partial^2_{pq}\mathcal W^{-1}_p(f)) +\Lamhda^{21}\Lambdq^{21}\paruial^2_{qq}\tilde{A} \partmzl^2_{pp}\matmeal F^{-1}_l(n))=a\alpka e^{-q}\partial^2_{pp}\majhcal F^{-1}_p(f)=\\ &&P^3(\tilde A,\mathcdl F^{-1}_'(f))=(-1)^6a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)=a\alpra e^{-q}\pattlal^3_{ppp}\mathcal S^{-1}_p(f)\emq{alifned}$$ and $P^k(\tilde A,\mathcal F^{-1}_p(f))=a\alphz e^{-q}\paruial^k_{p\dots p}\mathcak F^{-1}_p(f)\quad\forall k \geq 4,$ - Iv $\tipde A=b\beta e^q+dp.$ $$P^k(\hilde A,\mathxal S^{-1}_p(f))=(-1)^k b\beta e^q\oartial^k_{p\dots p}\mathcal F^{-1}_p(f)$$with $\quad\forall k \geq 2 $ - Iw $ \tikde{A}=a\alpha e^{-w}+b\bfja e^q+dp,$ $$\begii{alignvd} &&P^2(\tilde{A},\mathcal X^{-1}_p(f)) = \Lakbda^{12}\Lambda^{12}\partlal^2_{pp}\vildw{A} \partial^2_{qq}\mathcal F^{-1}_'(f)) + \Lambda^{12}\Lambda^{21}\pattial^2_{pq}\tilge{C} \partial^2_{qp}\mathcal F^{-1}_p(d)) + \\ &&\Lamtda^{21}\Lxnbdx^{12}\padtmal^2_{sp}\tildf{A} \'artial^2_{pq}\mafhcal F^{-1

partial^2_{qq}\mathcal F^{-1}_p(f)) + \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A} \partial^2_{qp}\mathcal F^{-1}_p(f)) + \partial^2_{pq}\mathcal +\Lambda^{21}\Lambda^{21}\partial^2_{qq}\tilde{A} \partial^2_{pp}\mathcal e^{-q}\partial^2_{pp}\mathcal F^{-1}_p(f)=\\ &&P^3(\tilde F^{-1}_p(f)\end{aligned}$$ $P^k(\tilde A,\mathcal F^{-1}_p(f))=a\alpha p}\mathcal F^{-1}_p(f)\quad\forall k 4,$ - If $\tilde A=b\beta e^q+dp.$ A,\mathcal F^{-1}_p(f))=(-1)^k b\beta e^q\partial^k_{p\dots p}\mathcal F^{-1}_p(f)$$with $\quad\forall k \geq 2 $ - If \tilde{A}=a\alpha e^{-q}+b\beta e^q+dp,$ $$\begin{aligned} &&P^2(\tilde{A},\mathcal F^{-1}_p(f)) = \Lambda^{12}\Lambda^{12}\partial^2_{pp}\tilde{A} \partial^2_{qq}\mathcal F^{-1}_p(f)) + \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A} \partial^2_{qp}\mathcal + &&\Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A} F^{-1

partial^2_{qq}\mathcal F^{-1}_p(f)) + \Lambda^{12}\lambda^{21}\partIal^2_{pq}\TilDe{A} \PaRtiaL^2_{qp}\mAthcal F^{-1}_p(f)) + \\ &&\LambdA^{21}\lambDa^{12}\partial^2_{qp}\tilde{A} \partiAl^2_{pq}\mAtHCal F^{-1}_P(F)) +\LAmbda^{21}\lambda^{21}\pARtIAL^2_{qq}\TiLdE{A} \pArTIaL^2_{pp}\maThcAl F^{-1}_p(f))=a\aLpha e^{-q}\partIal^2_{Pp}\Mathcal F^{-1}_p(f)=\\ &&P^3(\tILdE A,\mathcal F^{-1}_P(f))=(-1)^6a\Alpha e^{-q}\partiAl^3_{pPp}\mathCaL F^{-1}_p(F)=A\alphA e^{-q}\PartiAl^3_{ppp}\mAThcal F^{-1}_P(f)\end{aligNeD}$$ And $P^k(\tILde A,\matHCAl f^{-1}_p(f))=a\Alpha e^{-q}\partial^k_{p\dOTs P}\Mathcal F^{-1}_p(f)\quad\Forall K \gEQ 4,$ - IF $\TIldE A=b\Beta e^q+dp.$ $$P^k(\TiLde A,\mAThcal F^{-1}_p(F))=(-1)^K b\BETA e^q\PArtial^k_{p\dots p}\Mathcal F^{-1}_p(f)$$wITh $\qUad\forAlL k \gEQ 2 $ - If $ \tilDe{A}=a\aLpHA e^{-q}+B\beta e^q+dp,$ $$\beGin{aLigned} &&P^2(\tiLde{A},\maTHcal F^{-1}_p(f)) = \lAmbda^{12}\LaMbda^{12}\paRtiAl^2_{pP}\tilDE{A} \PaRtiAl^2_{QQ}\maTHcAl F^{-1}_P(F)) + \LaMbda^{12}\LambDa^{21}\PaRtial^2_{Pq}\tiLDE{a} \PartIal^2_{Qp}\maThcal f^{-1}_p(f)) + \\ &&\Lambda^{21}\LambDa^{12}\pArtiAL^2_{qp}\Tilde{a} \partIal^2_{pQ}\mAthcaL F^{-1

partial^2_{qq}\mathcal F^{ -1}_p(f)) + \ Lam bda ^{ 12}\ Lamb da^{21}\partia l ^2_{ pq}\tilde{A} \partial^ 2_{qp }\ m athc a lF^{-1 }_p(f)) + \ \ & &\L am b da ^{21} \La mbda^{1 2}\partial ^2_ {q p}\tilde{A}\ pa rtial^2_{p q}\ mathcal F^{- 1}_ p(f)) +\L a mbda^ {21 }\Lam bda^{2 1 }\part ial^2_{qq }\ t ilde{A } \parti a l ^2 _{pp }\mathcal F^{-1}_ p (f ) )=a\alpha e^{- q}\par ti a l^ 2 _ {pp }\m athcal F^{ -1 }_p(f ) =\\ && P ^ 3 (\t i lde A,\mathca l F^{-1}_p( f ))= (-1)^6 a\ alp h a e^{- q}\pa rt i al^ 3_{ppp}\mat hcal F^{-1}_p (f)=a\ a lpha e^ { -q}\par tial^3 _{p pp} \mat h ca lF^{ -1 } _p( f )\ end { ali gned}$$an d$P^k( \til d e A ,\ma thc al F ^{-1} _p(f))=a\alph a e ^{-q } \pa rtial ^k_{p \dot sp}\ma thcalF^{-1 }_ p(f)\quad\foral l k\geq 4,$ - If$\ tilde A=b\be tae^q +dp.$ $ $P^k(\t i lde A , \ m at hcal F^{-1}_p(f))= (- 1 ) ^k b\betae^q\pa r ti al ^ k_{p\dot sp}\ math c a l F^{ -1}_ p (f )$$with$\quad \ fo ra ll k \g eq 2 $ - I f $ \til d e{A} =a\alp ha e^{-q }+b\b e ta e^q+dp,$ $$ \ begin{aligned } & & P^2( \ti lde{A},\mat hcal F^{- 1}_p ( f) ) = \Lamb da^{1 2} \ La m bda^{12}\partial^2_ {p p}\til de{A} \partial^2_{ qq}\mathca l F ^{-1}_p( f))+\Lambda^{12}\L ambda ^{21}\part i al^2_{pq }\til de{A} \p artial^2_ { q p}\mathc alF^{ -1} _p( f ) ) + \\ &&\ L a mbda ^{ 21}\Lam bda ^{12}\p art ial ^2_ {qp }\ tilde{A}\partial ^2 _{ pq }\ mat hcalF ^{-1

partial^2_{qq}\mathcal F^{-1}_p(f))_ +_ \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A} \partial^2_{qp}\mathcal F^{-1}_p(f))_ +_\\ _ __&&\Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A} \partial^2_{pq}\mathcal F^{-1}_p(f))_ +\Lambda^{21}\Lambda^{21}\partial^2_{qq}\tilde{A}_\partial^2_{pp}\mathcal F^{-1}_p(f))=a\alpha e^{-q}\partial^2_{pp}\mathcal F^{-1}_p(f)=\\ _ _&&P^3(\tilde_A,\mathcal F^{-1}_p(f))=(-1)^6a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)=a\alpha e^{-q}\partial^3_{ppp}\mathcal F^{-1}_p(f)\end{aligned}$$ and $P^k(\tilde A,\mathcal F^{-1}_p(f))=a\alpha e^{-q}\partial^k_{p\dots p}\mathcal F^{-1}_p(f)\quad\forall k_\geq_4,$ - __If_$\tilde A=b\beta e^q+dp.$ $$P^k(\tilde A,\mathcal_F^{-1}_p(f))=(-1)^k b\beta e^q\partial^k_{p\dots p}\mathcal F^{-1}_p(f)$$with_$\quad\forall k_\geq 2 $ - If $ \tilde{A}=a\alpha_e^{-q}+b\beta_e^q+dp,$ $$\begin{aligned} _ &&P^2(\tilde{A},\mathcal F^{-1}_p(f)) = \Lambda^{12}\Lambda^{12}\partial^2_{pp}\tilde{A} \partial^2_{qq}\mathcal F^{-1}_p(f)) _+ \Lambda^{12}\Lambda^{21}\partial^2_{pq}\tilde{A} \partial^2_{qp}\mathcal F^{-1}_p(f)) _+ \\ ___&&\Lambda^{21}\Lambda^{12}\partial^2_{qp}\tilde{A} \partial^2_{pq}\mathcal F^{-1

first order $\alpha_{\rm SPH}$ term, while preventing particle interpenetration in the strong accretion shock. We have tested lowering the $\beta_{\rm SPH}$ value below the adopted value and found that it can cause numerical artefacts. Following @artymowicz96, we calculate the time scale on which the disc undergoes significant viscous evolution, $\tau_{\nu}=Re\,\Omega^{-1}$, assuming the Reynolds number, $Re$, and angular frequency, $\Omega$, are given by: $$\label{eq:reynolds_eda1} Re = \frac{(r/\langle h \rangle)^2}{\alpha_{\nu}}\\ \Omega = \sqrt{\frac{GM_{*}}{r^3}}\\$$ with $r$ the radial distance to the star and $\langle h \rangle$ the (resolution-dependent) average smoothing length in the disc at that radius. For a simulation of 16.000 disc particles, and assuming the corresponding $\alpha_{\nu}$, this gives us a viscous time scale, $\tau_{\nu}$, of roughly 10.000 years at 10 AU. We can also calculate the physical viscous time scale at 10 AU by replacing $\langle h \rangle$ with the scale height of the disc, $H=\frac{c_s}{\Omega}$, at that radius. This leads to a time scale of $3\times10^5$ years. In order to remain safely in the regime where the simulation is not dominated by viscous evolution of the disc, we evolve the whole set-up for 2500 years and start the inflow at $t=0$. The numerical viscosity in the disc changes during the simulations with Gadget2 and we will discuss the numerical effects in Sec. \[sec:convergence\_eda1\] and \[sec:consistency\_eda1\]. Distinguishing the disc from the ISM {#sec:disc_recognition_eda1} ------------------------------------ \ \ To determine how much ISM has been swept-up by the disc at any moment in time, we need to differentiate between the disc and the ISM flow. This is not straightforward since there is no clear separation between the outer edge of the disc and the flow. In order to distinguish between the disc and the ISM flow, we use a clump finding algorithm to identify different groups in the parameter space of the angular speed ($v_{\theta}$), the density ($\rho$) and the speed along the axis of the cylinder($v_

first order $ \alpha_{\rm SPH}$ term, while preventing atom interpenetration in the impregnable accretion shock. We have tested turn down the $ \beta_{\rm SPH}$ value below the adopt value and found that it can cause numeral artifact. Following @artymowicz96, we calculate the time scale on which the disc undergo significant viscous evolution, $ \tau_{\nu}=Re\,\Omega^{-1}$, assuming the Reynolds numeral, $ Re$, and angular frequency, $ \Omega$, are given by: $ $ \label{eq: reynolds_eda1 } Re = \frac{(r/\langle h \rangle)^2}{\alpha_{\nu}}\\ \Omega = \sqrt{\frac{GM_{*}}{r^3}}\\$$ with $ r$ the radial distance to the star and $ \langle h \rangle$ the (resolution - dependent) modal smoothing length in the disc at that radius. For a pretense of 16.000 disc particles, and assuming the corresponding $ \alpha_{\nu}$, this pass us a viscous time scale, $ \tau_{\nu}$, of roughly 10.000 years at 10 AU. We can besides calculate the physical viscous time scale at 10 AU by replacing $ \langle planck's constant \rangle$ with the scale height of the disc, $ H=\frac{c_s}{\Omega}$, at that radius. This leads to a time scale of $ 3\times10 ^ 5 $ years. In order to remain safely in the government where the simulation is not dominated by gluey development of the disk, we evolve the whole hardening - up for 2500 years and start the inflow at $ t=0$. The numerical viscosity in the disc change during the simulations with Gadget2 and we will discuss the numerical effects in Sec. \[sec: convergence\_eda1\ ] and \[sec: consistency\_eda1\ ]. Distinguishing the disc from the ISM { # sec: disc_recognition_eda1 } ------------------------------------ \ \ To decide how much ISM has been swept - up by the disc at any moment in time, we necessitate to differentiate between the phonograph record and the ISM flow. This is not straightforward since there is no clear separation between the out edge of the disc and the flow. In decree to distinguish between the disc and the ISM flow, we use a ball finding algorithm to identify unlike groups in the parameter space of the angular speed ($ v_{\theta}$), the concentration ($ \rho$) and the speed along the axis of the cylinder($v _

figst order $\alpha_{\rm SPH}$ tevm, while preventnbg parvicle ihterpenegration in the strong accretmon whock. We have tested lowerivg the $\bena_{\rm SPH}$ calut below the adoptxs value and fkmnd tkav it can cause kumerical astefacts. Followhne @crtymowicz96, we calculate the time scaje on wnifh the disc unqergpqs sjgnificant viscous evolution, $\tau_{\nu}=De\,\Omega^{-1}$, assuming the Reynolds number, $Re$, and anhulag frequency, $\Omega$, wre given bi: $$\lafwl{eq:reynolds_dda1} Re = \frab{(x/\langle h \rzngle)^2}{\alpha_{\nu}}\\ \Omega = \sqrt{\frac{GM_{*}}{r^3}}\\$$ dith $x$ the radiao eishdnce to the star and $\langle m \rangla$ the (rrsolution-depencenv) avwrage smoothing lengti in the disc at thaj radius. Fmr a simulation of 16.000 dusc pdrtiwles, qnd asaukihg the cocresponding $\alpha_{\nu}$, thus gives us a viscois nome scale, $\tah_{\nu}$, of rjughly 10.000 years at 10 AU. We can also calculdte the physical viscous tume scale at 10 AU by rgplacing $\lwngle h \rangle$ with the scale height of the disc, $V=\frac{r_s}{\Umeyq}$, at gyah radius. This leads to a time scale of $3\times10^5$ rsats. In order to rcmain safely in thr gebyme where the simulcfikn is not dominatef by vifcous evolutiog of the disc, we evolve the whooe set-up for 2500 years and start tke inflow at $t=0$. Tne nukerical viscosity in thz disc changes duging the akmulations with Eadnet2 and we will discuss the ntmerical xffeccs in Sez. \[sev:convewgence\_eda1\] wnd \[scw:consistency\_eda1\]. Dishingunshinc the disc from the ISM {#sec:disc_recognitioi_xda1} ------------------------------------ \ \ To determime hof much IFM hax been swept-u[ by the disc ct any mjment in time, wv need to differentiwte between tvg disc and thx ISM flor. Thus iw not sgfaightforward xince thege is bo clear separatiok betdsen the outer ebye of the disc anc tfe sllw. Ig order to divtineuirn betdeen tht dlsc and the ISM flow, we use a cmump finding algorotmm to idebtify disferent groupx in the parameter spacx of tie angilat speed ($v_{\theta}$), the density ($\rho$) znd the soeeq along the wxis of the cyliuder($v_

first order $\alpha_{\rm SPH}$ term, while preventing in strong accretion We have tested below adopted value and that it can numerical artefacts. Following @artymowicz96, we calculate time scale on which the disc undergoes significant viscous evolution, $\tau_{\nu}=Re\,\Omega^{-1}$, assuming the number, $Re$, and angular frequency, $\Omega$, are given by: $$\label{eq:reynolds_eda1} Re = \frac{(r/\langle \rangle)^2}{\alpha_{\nu}}\\ = with the radial distance to the star and $\langle h \rangle$ the (resolution-dependent) average smoothing length in disc at that radius. For a simulation of disc particles, and assuming corresponding $\alpha_{\nu}$, this gives us viscous scale, $\tau_{\nu}$, roughly years 10 AU. We also calculate the physical viscous time scale at 10 AU by replacing $\langle h \rangle$ with the height of $H=\frac{c_s}{\Omega}$, at radius. leads a time scale years. In order to remain safely where the simulation is not dominated by viscous of the we evolve the whole set-up for years and start the inflow at $t=0$. The viscosity in the disc changes during the simulations with Gadget2 and we will discuss the in Sec. \[sec:convergence\_eda1\] and Distinguishing the disc the {#sec:disc_recognition_eda1} \ To determine much ISM has been swept-up by the disc at any moment time, we need to differentiate between the disc and the This not straightforward since is no clear separation the edge of the disc flow. order the and ISM flow, we use clump finding algorithm to identify groups in the parameter ($v_{\theta}$), the density ($\rho$) and the speed along axis of the cylinder($v_

first order $\alpha_{\rm SPH}$ term, wHile prevenTing pArtIclE iNterPeneTration in the stROng aCcretion shock. We have tesTed loWeRIng tHE $\bEta_{\rm sPH}$ valuE BeLOW thE aDoPteD vALuE and fOunD that it Can cause nuMerIcAl artefacts. FOLlOwing @artymOwiCz96, we calculatE thE time sCaLe oN Which The Disc uNdergoES signiFicant visCoUS evoluTIon, $\tau_{\nU}=rE\,\OMega^{-1}$, Assuming the ReynolDS nUMber, $Re$, and angulAr freqUeNCy, $\oMEga$, Are Given by: $$\labEl{Eq:reyNOlds_eda1} rE = \fRAC{(R/\laNGle h \rangle)^2}{\alpHa_{\nu}}\\ \Omega = \sqRT{\frAc{GM_{*}}{r^3}}\\$$ wItH $r$ tHE radiaL distAnCE to The star and $\lAnglE h \rangle$ tHe (resoLUtion-dePEndent) aVerage SmoOthIng lENgTh In tHe DIsc AT tHat RAdiUs. For a siMuLaTion oF 16.000 disC PARTiclEs, aNd asSuminG the corresponDinG $\alpHA_{\nu}$, This gIves uS a viScOus tiMe scalE, $\tau_{\nU}$, oF roughly 10.000 years at 10 aU. We Can also caLcuLaTe tHe PhysiCAl viscOus TimE scale aT 10 AU by rePLacInG $\LANgLe h \rangle$ with the scAlE HEiGht of the Disc, $H=\fRAc{C_s}{\oMega}$, at thAt RadIus. THIS leadS to a TImE scale of $3\Times10^5$ yEArS. IN order tO rEmain sAfEly In tHe regIMe whEre the SimulatiOn is nOT dominated by viSCous evolution OF tHE DiSC, we eVolVe the whole sEt-up FOr 2500 yeArs aND sTarT The inFlow aT $t=0$. tHe NUmerical viscosity in ThE disc cHangeS during the simUlations wiTH gAdget2 and We wiLL dIScuss the numeriCal efFects in Sec. \[SEc:converGence\_Eda1\] and \[seC:consisteNCY\_eda1\]. DistIngUisHinG thE DIsC from the ISM {#seC:DIsc_rEcOgnitioN_edA1} ------------------------------------ \ \ To deteRmiNe hOw mUch iSm has been sWept-up by ThE dIsC aT anY momeNT in time, wE nEed To DifFerenTIate beTween The dIsC aND thE ISM floW. thIS Is noT sTrAighTfoRwArd siNce tHEre Is no cleAr separatIon BEtweEn ThE outer eDge of the disc aNd The flow. In oRdEr tO distiNGUish betwEen the disc and the ISM flow, WE use a clUmp FindiNg alGorithm to IdeNtify dIffERent grOups in The paRaMetER Space OF ThE anGuLar speed ($v_{\tHETa}$), tHe denSiTy ($\rhO$) and the Speed along the axis oF The Cylinder($v_

first order $\alpha_{\rmSPH}$ term , whi lepre ve ntin g pa rticle interpe n etra tion in the strong acc retio ns hock . W e hav e teste d l o w eri ng t he$\ b et a_{\r m S PH}$ va lue belowthe a dopted value an d found th atit can cause nu merica lart e facts . Follo wing @ a rtymow icz96, we c a lculat e the ti m e s cale on which the dis c u n dergoes signif icantvi s co u s ev olu tion, $\ta u_ {\nu} = Re\,\Om e ga ^ { - 1}$ , assuming the Reynolds n u mbe r, $Re $, an d angul ar fr eq u enc y, $\Omega$ , ar e given b y: $$\ l abel{eq : reynold s_eda1 } R e = \fr a c{ (r /\l an g leh \ ran g le) ^2}{\alp ha _{ \nu}} \\ \ O m e g a =\sq rt{\ frac{ GM_{*}}{r^3}} \\$ $ wi t h $ r$ th e rad ialdi stanc e to t he st ar and $\langle h \ra ngle$ the (r es olu ti on-de p endent ) a ver age smo othingl eng th i n t he disc at that ra di u s .For a si mulati o nof 16.000 d is c p arti c l es, a nd a s su ming the corre s po nd ing $\a lp ha_{\n u} $,thi s giv e s us a vis cous tim e sca l e, $\tau_{\nu} $ , of roughly1 0. 0 0 0y ears at 10 AU. Wecana lsocalc u la tet he ph ysica lv is c ous time scale at 1 0AU byrepla cing $\langle h \rangle $ w ith thescal e h e ight of the di sc, $ H=\frac{c_ s }{\Omega }$, a t that r adius. Th i s leads t o a ti mesca l e o f $3\times10^ 5 $ yea rs . In or der to rem ain sa fel y i nthe regim e whereth esi mu lat ion i s not dom in ate dbyvisco u s evol ution ofth ed isc , we ev o lv e thewh ol e se t-u pfor 2 500y ear s and s tart theinf l ow a t$t =0$. Th e numerical v is cosity inth e d isc ch a n ges duri ng the simulations with Gadget2 an d wewill discussthe numer ica l effec ts inSec.\[ sec : c onver g e nc e\_ ed a1\] and \ [ s ec: consi st ency \_eda1\ ]. Distinguishing the disc from th e I SM { # s ec :di s c_ r eco gn i tio n _ eda1} --------- ---------- -- - -- ---------- - - \ \ Todetermi ne ho w much I SM has be en swept- up byt h e d isc at any momentin time,w e nee d t o dif fer entiat ebet weenthe di s c a nd th e ISMfl ow. Th is is n ot strai ghtforward since thereis noclear se parationbet w een the oute r ed ge of thedis c a nd th e f l ow. I n or d er to disti ngui s h between th e d i s cand the ISM f l ow, we u sea clump fin ding algorithm to identify diffe rent g rou psi n th eparameter spac e o ft h e angula rspeed ($v_{ \theta}$ ), the d ensity ($\rh o$) and t he speedalon g t he axis o f t he cylinde r( $v _

first_order $\alpha_{\rm_SPH}$ term, while preventing_particle interpenetration_in_the strong_accretion_shock. We have_tested lowering the_$\beta_{\rm SPH}$ value below_the adopted value_and_found that it can cause numerical artefacts. Following @artymowicz96, we calculate the time scale on_which_the disc_undergoes_significant_viscous evolution, $\tau_{\nu}=Re\,\Omega^{-1}$, assuming the_Reynolds number, $Re$, and angular_frequency, $\Omega$,_are given by: $$\label{eq:reynolds_eda1} Re = \frac{(r/\langle h \rangle)^2}{\alpha_{\nu}}\\ \Omega_=_\sqrt{\frac{GM_{*}}{r^3}}\\$$ with $r$_the radial distance to the star and $\langle h_\rangle$ the (resolution-dependent) average smoothing length_in the disc_at_that_radius. For a simulation_of 16.000 disc particles, and assuming_the corresponding $\alpha_{\nu}$, this gives us_a viscous time scale, $\tau_{\nu}$, of roughly_10.000 years at 10 AU. We_can also calculate the physical_viscous time_scale at 10 AU by_replacing $\langle h_\rangle$ with_the scale height_of the disc, $H=\frac{c_s}{\Omega}$, at that_radius. This leads_to a time scale of $3\times10^5$_years._In order to_remain_safely_in the_regime where the_simulation_is not_dominated_by viscous evolution of the disc,_we_evolve the whole set-up for 2500 years_and start the inflow_at_$t=0$. The numerical viscosity_in the disc changes during_the simulations with Gadget2 and we_will discuss_the numerical_effects in Sec. \[sec:convergence\_eda1\] and \[sec:consistency\_eda1\]. Distinguishing the disc from the ISM_{#sec:disc_recognition_eda1} ------------------------------------ \ \ To determine how much ISM has_been swept-up by the_disc at_any_moment in time,_we_need to_differentiate between the disc and the ISM_flow. This_is not straightforward since there is_no clear separation between_the_outer edge of the disc and_the flow. In order to distinguish_between the disc and the_ISM_flow,_we use a clump finding_algorithm to identify different groups in_the parameter space_of the angular speed ($v_{\theta}$), the density_($\rho$)_and the speed along the axis_of_the cylinder($v_

{eq:diffusion_energy} E_{\rm D} \equiv E_0 = \frac{\langle \phi_0 | \hat{H} | \Psi_{\rm T} \rangle}{\langle \phi_0 | \Psi_{\rm T} \rangle} & = & \frac{\int f({\bf R}) E_{\rm L}({\bf R}) \, d{\bf R}}{\int f({\bf R}) \, d{\bf R}} \\ & \simeq & \frac{1}{M} \sum_{i=1}^M E_{\rm L}({\bf R}_i) \;.\end{aligned}$$ Some example DMC data are shown in figure \[fig:graphit\_silane\]. ![DMC data for a silane (SiH$_4$) molecule, with the ions represented by pseudopotentials. The upper panel shows the fluctuations in the population of configurations arising from the branching process used to simulate equation (\[eq:short time GB\]). The reference energy, $E_{\rm T}$, is altered during the run to control the population. Specifically, the reference energy is set to return the population to the target population (128,000 configurations) on the same time-scale as the autocorrelation period of the energy data [@foulkes_2001]. The total energy is shown in the lower panel as a function of the move number. The black line shows the instantaneous value of the local energy averaged over the current population of configurations, the red line is the reference energy $E_{\rm T}$ and the green line is the best estimate of the DMC energy as the simulation progresses. The configurations at move number zero are from the output of a VMC simulation, and the energy decays rapidly from its initial VMC value of about -6.250 a.u. and reaches a plateau with a DMC energy of about -6.305 a.u. The data up to move 1000 are deemed to form the equilibration phase, and are discarded. []{data-label="fig:graphit_silane"}](graphit_silane){width="100.00000%"} Trial wave functions {#sec:psi_trial} ==================== Trial wave functions are of central importance in VMC and DMC calculations because they introduce importance sampling and control both the statistical efficiency and accuracy obtained. The accuracy of a DMC calculation depends on the nodal surface of the trial wave function via the fixed-node approximation, while in V

{ eq: diffusion_energy } E_{\rm D } \equiv E_0 = \frac{\langle \phi_0 | \hat{H } | \Psi_{\rm T } \rangle}{\langle \phi_0 | \Psi_{\rm T } \rangle } & = & \frac{\int f({\bf R }) E_{\rm L}({\bf R }) \, d{\bf R}}{\int f({\bf R }) \, d{\bf R } } \\ & \simeq & \frac{1}{M } \sum_{i=1}^M E_{\rm L}({\bf R}_i) \;.\end{aligned}$$ Some example DMC data are shown in name \[fig: graphit\_silane\ ]. ! [ DMC datum for a silane (SiH$_4 $) molecule, with the ions represent by pseudopotentials. The upper gore shows the fluctuations in the population of shape arise from the branching procedure used to simulate equality (\[eq: short time GB\ ]). The reference energy, $ E_{\rm T}$, is interpolate during the run to see the population. Specifically, the reference energy is sic to return the population to the target population (128,000 configurations) on the like time - scale as the autocorrelation time period of the energy datum [ @foulkes_2001 ]. The total department of energy is shown in the lower panel as a function of the motion number. The black line shows the instantaneous value of the local energy average over the current population of configurations, the red line is the reference energy $ E_{\rm T}$ and the green line is the best estimate of the DMC energy as the model progresses. The configurations at motion numeral zero are from the output of a VMC simulation, and the energy decay rapidly from its initial VMC value of about -6.250 a.u.   and reaches a plateau with a DMC energy of about -6.305 a.u. The datum up to move 1000 are deemed to form the equilibration phase, and are discarded. [ ] { data - label="fig: graphit_silane"}](graphit_silane){width="100.00000% " } Trial wave functions { # sec: psi_trial } = = = = = = = = = = = = = = = = = = = = Trial wave functions are of central importance in VMC and DMC calculations because they introduce importance sampling and see both the statistical efficiency and accuracy obtained. The accuracy of a DMC calculation depend on the nodal surface of the trial wave function via the fixed - node estimate, while in V

{eq:dlffusion_energy} E_{\rm D} \equiy E_0 = \frac{\langle \khu_0 | \hat{I} | \Psi_{\rj T} \rangld}{\langle \phi_0 | \Psi_{\rm T} \rangle} & = & \frac{\unt f({\bf R}) E_{\rm L}({\bf R}) \, d{\bw R}}{\int f({\bv R}) \, d{\bf R}} \\ & \wimeq & \frac{1}{M} \sum_{i=1}^M C_{\xm L}({\br R}_i) \;.\znv{aligned}$$ Some ewample DMC gata are shown iv yigure \[fig:graphit\_silane\]. ![DMC data for w silanr (DiH$_4$) molecule, wyth uhe iona represented by pseudopotentials. Fhe upptr panel shows the fluctuations in the populwtioj of configurationd arising ftkm eye branching process used to simulzte equation (\[eq:short time GB\]). Thd refzrence enerty, $E_{\rl T}$, is alterxd durpng the run to contron the pppulation. Speclficanly, the reference energy is set to return jhe populadikn to the target populathon (128,000 conwugufatjois) kn the saje time-scame as the aytocorrelation peripd if the energy data [@sotlkes_2001]. The total energy is shown in the nowsr panel as a function if the move number. Thg black lige shows the instantaneous value of the local enesgy atefagtd over rhf current population of configurations, the req lonv is the referencc energy $E_{\rm T}$ and tje dreen line is the bzat estimate of the DLC enerdy as the simujatipn progresses. The configurarions at movv nunber zero are from the output of s VMC simulation, and the enexgy dedays rapidlj from ita initial VMC valje pf about -6.250 a.u. and reaches a plwteau witi a DKC enerey og aboue -6.305 a.u. The fata m[ to move 1000 are deeled tl xorm the eeuilibration phase, and are discecded. []{data-labek="fhg:ggaphit_silcne"}](graihit_silane){width="100.00000%"} Erial wave funetions {#szc:psi_tfial} ==================== Trial save fuictions are jf central im[lrtance in VKC and DIC cqlcuoations cecause they imtroduce pmkortance sqmpling and controk bufh the statistiecl efficiency and aczurwcj outainqg. The accurawy ow a CMC cxlculation beoendx on the nodal surfawe or the trial wave fingtion via the fixqd-node approxomation, while in V

{eq:diffusion_energy} E_{\rm D} \equiv E_0 = \frac{\langle \hat{H} \Psi_{\rm T} \phi_0 | \Psi_{\rm \frac{\int R}) E_{\rm L}({\bf \, d{\bf R}}{\int R}) \, d{\bf R}} \\ & & \frac{1}{M} \sum_{i=1}^M E_{\rm L}({\bf R}_i) \;.\end{aligned}$$ Some example DMC data are shown figure \[fig:graphit\_silane\]. ![DMC data for a silane (SiH$_4$) molecule, with the ions represented pseudopotentials. upper shows fluctuations in the population of configurations arising from the branching process used to simulate equation (\[eq:short GB\]). The reference energy, $E_{\rm T}$, is altered the run to control population. Specifically, the reference energy set return the to target (128,000 configurations) on same time-scale as the autocorrelation period of the energy data [@foulkes_2001]. The total energy is shown in lower panel function of move The line shows the of the local energy averaged over of configurations, the red line is the reference $E_{\rm T}$ the green line is the best of the DMC energy as the simulation progresses. configurations at move number zero are from the output of a VMC simulation, and the rapidly from its initial value of about a.u. reaches plateau a DMC of about -6.305 a.u. The data up to move 1000 are to form the equilibration phase, and are discarded. []{data-label="fig:graphit_silane"}](graphit_silane){width="100.00000%"} Trial {#sec:psi_trial} Trial wave functions of central importance in and calculations because they introduce and both and obtained. accuracy of a DMC depends on the nodal surface the trial wave function in V

{eq:diffusion_energy} E_{\rm D} \equiV E_0 = \frac{\langLe \phi_0 | \Hat{h} | \PsI_{\rM T} \raNgle}{\Langle \phi_0 | \Psi_{\rm t} \RangLe} & = & \frac{\int f({\bf R}) E_{\rm L}({\bf R}) \, d{\bF R}}{\int F({\bF r}) \, d{\bf r}} \\ & \SiMeq & \frAc{1}{M} \sum_{i=1}^m e_{\rM l}({\Bf R}_I) \;.\eNd{AliGnED}$$ SOme exAmpLe DMC daTa are shown In fIgUre \[fig:graphiT\_SiLane\]. ![DMC datA foR a silane (SiH$_4$) mOleCule, wiTh The IOns rePreSenteD by pseUDopoteNtials. The UpPEr paneL Shows thE FLuCtuaTions in the populatIOn OF configurationS arisiNg FRoM THe bRanChing proceSs Used tO SimulatE EqUATIon (\[EQ:short time GB\]). THe reference ENerGy, $E_{\rm T}$, Is AltERed durIng thE rUN to Control the pOpulAtion. SpecIficalLY, the refERence enErgy is Set To rEturN ThE pOpuLaTIon TO tHe tARgeT populatIoN (128,000 cOnfigUratIONS) On thE saMe tiMe-scaLe as the autocoRreLatiON peRiod oF the eNergY dAta [@foUlkes_2001]. THe totAl Energy is shown in The lOwer panel As a FuNctIoN of thE Move nuMbeR. ThE black lIne showS The InSTANtAneous value of the loCaL ENeRgy averaGed oveR ThE cURrent popUlAtiOn of CONfiguRatiONs, The red liNe is thE ReFeRence enErGy $E_{\rm T}$ AnD thE grEen liNE is tHe best Estimate Of the dmC energy as the sIMulation progrESsES. thE ConfIguRations at moVe nuMBer zEro aRE fRom THe outPut of A Vmc sIMulation, and the energY dEcays rApidlY from its initiAl VMC value OF ABout -6.250 a.u. anD reaCHeS A plateau with a DmC eneRgy of about -6.305 A.U. The data Up to mOve 1000 are deEmed to forM THe equiliBraTioN phAse, AND aRe discarded. []{daTA-LabeL="fIg:graphIt_sIlane"}](grAphIt_sIlaNe){wIdTh="100.00000%"} Trial waVe functiOnS {#sEc:PsI_trIal} ==================== TrIAl wave fuNcTioNs Are Of cenTRal impOrtanCe in vMc aND DMc calculATiONS becAuSe They IntRoDuce iMporTAncE sampliNg and contRol BOth tHe StAtisticAl efficiency aNd Accuracy obTaIneD. The acCURacy of a DmC calculation depends on tHE nodal sUrfAce of The tRial wave fUncTion viA thE Fixed-nOde appRoximAtIon, WHIle in v

{eq:diffusion_energy} E_{\ rm D} \equ iv E_ 0 = \f ra c{\l angl e \phi_0 | \ha t {H}| \Psi_{\rm T} \rangle }{\la ng l e \p h i_ 0 | \ Psi_{\r m T } \ra ng le } & = &\frac {\i nt f({\ bf R}) E_{ \rm L }({\bf R}) \ , d {\bf R}}{\ int f({\bf R})\,d{\bfR} } \ \ & \s ime q & \ frac{1 } {M} \s um_{i=1}^ ME _{\rmL }({\bfR } _i ) \; .\end{aligned}$$S om e example DMC d ata ar es ho w n in fi gure \[fig :g raphi t \_silan e \] . ![D M C data for asilane (SiH $ _4$ ) mole cu le, with t he io ns rep resented by pse udopotent ials.T he uppe r panelshowsthe fl uctu a ti on s i nt hep op ula t ion of conf ig ur ation s ar i s i n g fr omthebranc hing processuse d to sim ulate equa tion ( \[eq: shorttimeGB \]). The refere nceenergy, $ E_{ \r m T }$ , isa ltered du rin g the r un to c o ntr ol t h epopulation. Specif ic a l ly , the re ferenc e e ne r gy is se ttoretu r n thepopu l at ion to t he tar g et p opulati on (128, 00 0 c onf igura t ions ) on t he sametime- s cale as the au t ocorrelationp er i o do f th e e nergy data[@fo u lkes _200 1 ]. Th e tota l ene rg y i s shown in the lower p anel a s a f unction of th e move num b e r . The bl ackl in e shows the ins tanta neous valu e of thelocal energyaveragedo v er the c urr ent po pul a t io n of configur a t ions ,the red li ne is t heref ere nce e nergy $E_ {\rm T}$ a nd t he gr een l i ne is th ebes test imate of the DMCener gy a s th e simul a ti o n pro gr es ses. Th econfi gura t ion s at mo ve number ze r o ar efr om theoutput of a V MC simulatio n, an d thee n ergy dec ays rapidly from its in i tial VM C v alueof a bout -6.2 50a.u. a ndr eaches a pla teauwi tha DMC e n e rg y o fabout -6.3 0 5 a. u. Th edata up tomove 1000 are deem e d t o form the eq uil ibra t i on ph a se , an da red i scarded. []{dat a-label="f ig : gr aphit_sila n e"} ]( graphit _silane ){wid t h="100. 00000%"} Trial wa ve fun c t ion s {#sec:ps i_trial} ======== = ===== = == === Tri al wav efun ction s areo f c entra l impo rt ance i n VMC a nd DMC c alculations because the y intr oduce im portancesam p lin g and con trol both thesta tis tical ef f icien cy a n dacc u racyobta i ned. Thea cc ura c y o f a DMC cal c u l ati on de pen d s on t he n odal surface of t h e trial wave f unct i o n v iat he f ix ed-node approx ima ti o n , whilein V

{eq:diffusion_energy} E_{\rm D}_\equiv E_0_= \frac{\langle \phi_0 |_\hat{H} |_\Psi_{\rm_T} \rangle}{\langle \phi_0_|_\Psi_{\rm T} \rangle}_& = &_\frac{\int f({\bf R}) E_{\rm L}({\bf_R}) \, d{\bf_R}}{\int_f({\bf R}) \, d{\bf R}} \\ & \simeq & \frac{1}{M} \sum_{i=1}^M E_{\rm L}({\bf R}_i) \;.\end{aligned}$$_Some_example DMC_data_are_shown in figure \[fig:graphit\_silane\]. ![DMC data_for a silane (SiH$_4$) molecule,_with the_ions represented by pseudopotentials. The upper panel shows_the_fluctuations in the_population of configurations arising from the branching process used_to simulate equation (\[eq:short time GB\])._The reference energy,_$E_{\rm_T}$,_is altered during the_run to control the population. Specifically,_the reference energy is set to_return the population to the target population_(128,000 configurations) on the same time-scale_as the autocorrelation period of_the energy_data [@foulkes_2001]. The total energy_is shown in_the lower_panel as a_function of the move number. The_black line shows_the instantaneous value of the local_energy_averaged over the_current_population_of configurations,_the red line_is_the reference_energy_$E_{\rm T}$ and the green line_is_the best estimate of the DMC energy_as the simulation progresses._The_configurations at move number_zero are from the output_of a VMC simulation, and the_energy decays_rapidly from_its initial VMC value of about -6.250 a.u. and reaches a plateau_with a DMC energy of about_-6.305 a.u. The data_up to_move_1000 are deemed_to_form the_equilibration phase, and are discarded. []{data-label="fig:graphit_silane"}](graphit_silane){width="100.00000%"} Trial wave_functions {#sec:psi_trial} ==================== Trial_wave functions are of central importance_in VMC and DMC_calculations_because they introduce importance sampling and_control both the statistical efficiency and_accuracy obtained. The accuracy of_a_DMC_calculation depends on the nodal_surface of the trial wave function_via the fixed-node_approximation, while in V

can be seen in Fig.3 that the transmission distance decreases with the dimension $N$. Near the falling critical point, the probability of the wrong successful output is comparable to that of the right successful output, as $N$ grows larger, the proportion occupied by the wrong successful output becomes more significant, as well as the noise entropy. This causes a decrease in the transmission. In summary, we have proposed an experimental scheme to implement high-dimensional MDI-QKD by exploiting the discrete position basis and momentum basis as two MUBs. It has the same level of security as that of polarization-based two-dimensional MDI-QKD. Given a high enough dimension for this experimental proposal, not only the quantum secret key rate can be enlarged significantly, but also that the quantum-Nit error rate coming from the overlap between diffraction peaks of different basis states in X basis measurement will be reduced substantially. Furthermore, the transmission distance decreases as an increasing in dimension, but this drawback can be overcome by the method of quantum relay. A combination of high-dimensional MDI-QKD and existed optimization methods would serve as a rational approach for further improvement of secret key generation rate. This work is supported by Young fund of Jiangsu Natural Science Foundation of China (SJ216025), National fund incubation project (NY217024), Scientific Research Foundation of Nanjing University of Posts and Telecommunications (NY215034), National Natural Science Foundations of China (Grant No. 61475075). --- abstract: | The rapid involution of the mobile generation with incipient data networking capabilities and utilization has exponentially increased the data traffic volumes. Such traffic drains various key issues in 5G mobile backhaul networks. Security of mobile backhaul is of utmost importance; however, there are a limited number of articles, which have explored such a requirement. This paper discusses the potential design issues and key challenges of the secure 5G mobile backhaul architecture. The comparisons of the existing state-of-the-art solutions for secure mobile backhaul, together with their major contributions have been explored. Furthermore, the paper discussed various key issues related to Quality of Service (QoS), routing and scheduling, resource management, capacity enhancement, latency, security-management, and handovers using mechanisms like Software Defined Networking and millimeter Wave technologies. Moreover, the trails of research challenges and future directions are additionally presented.\ \ **Keywords: 5G, Mobile Backhaul, Security

can be seen in Fig.3 that the transmission distance decreases with the dimension $ N$. Near the accrue critical item, the probability of the wrong successful output is comparable to that of the right successful output signal, as $ N$ grows larger, the symmetry occupied by the wrong successful output become more significant, as well as the randomness entropy. This causes a decrease in the transmission. In summary, we have proposed an experimental schema to implement high - dimensional MDI - QKD by exploiting the discrete position basis and momentum footing as two MUBs. It has the same level of security system as that of polarization - based two - dimensional MDI - QKD. Given a high enough dimension for this experimental marriage proposal, not only the quantum privy key rate can be enlarged importantly, but also that the quantum - Nit error rate coming from the overlap between diffraction peaks of unlike footing department of state in X basis measurement will be reduced substantially. Furthermore, the transmission distance decreases as an increasing in dimension, but this drawback can be get the better of by the method of quantum relay. A combination of high - dimensional MDI - QKD and exist optimization method would serve as a rational approach for further improvement of unavowed cardinal generation rate. This work is corroborate by Young fund of Jiangsu Natural Science Foundation of China (SJ216025), National fund incubation project (NY217024), Scientific Research Foundation of Nanjing University of Posts and Telecommunications (NY215034), National Natural Science Foundations of China (Grant No. 61475075). --- abstract: | The rapid involution of the mobile generation with incipient data networking capabilities and utilization has exponentially increased the datum traffic volume. Such dealings drain various key issues in 5 gravitational constant mobile backhaul networks. Security of fluid backhaul is of utmost importance; however, there be a limited number of article, which have explored such a prerequisite. This paper discusses the likely design issues and key challenges of the impregnable 5 G fluid backhaul architecture. The comparisons of the existing department of state - of - the - artwork solutions for secure mobile backhaul, in concert with their major contributions have been explored. Furthermore, the paper discussed versatile cardinal issues related to Quality of Service (QoS), rout and scheduling, resource management, capacity enhancement, latency, security - management, and handovers using mechanisms like Software Defined Networking and millimeter Wave technologies. furthermore, the trails of research challenges and future directions are additionally presented.\ \ * * Keywords: 5 G, Mobile Backhaul, Security

caj be seen in Fig.3 that tht transmission distance vecreasss with ghe dimension $N$. Near the falpibg crutical point, the probacility of the wrobg snccessful output is comparable bo thct of the right xuccessful output, as $N$ gsods larger, the proportion occupied by ehe wromg successful oujput nqcomss more significant, as well as the noise tntropy. This causex a decrease in the transmlssiln. In summary, we hage proposed an qzperimental rcheme to pkplement hjgh-dimensional MDI-QKD by exploiging che discretg lodhtion basis and iomentum baslx as tfo MUBs. It has the sake nevwl of security as thav of polarization-basgd two-dimetsnonal MDI-QKD. Given a yith enmugh dimdbsiun roc tgis exoermmental prolosal, not obly the quantum sectee key rate can be enjawged significantly, but also that the qudnthm-Nit error rate coming from the overlap betaeen diffwaction peaks of different basis states in X basiv meaajrenekt wkol be reduced substantially. Furthermore, the tragamosxion distance becreases as an onfrrwsing in dimevsion, yht this drawback can be ovetcome vy the meuhod pf quantum relay. A combinatuon of high-dpmenwional MDI-QKD and zxisted optioizajion mrthods would serve as a ratjonal approwch for fhfther improvemeng on sacret key generation rate. Tris work ms su'ported cy Ypung ftnd of Jiajgsu Kdtural Science Foujdatiln of China (DJ216025), National fund incubation prooxct (NY217024), Scientofhc Gesearch Youndabion of Nanjing University of Posts cnd Teuecommunicztions (IY215034), National Gatural Scienwg Foundations of Chinw (Grqnt Bo. 61475075). --- absgfact: | The rspid involution of rhe mobile generatlon wkfh incipient dacc betworking capanilktiqs aid utynization has expunevyiallh increased ghe cata traffic volumes. Sucg traffic drains vsrlous key ussues ig 5G mobile bavkhaul networks. Sefuritb of mubile baskhaul is of utmost importance; however, hheve are a limieed kumbgr of artickes, which have explored such a requiremxnt. This paper discussgs the potential desiyn issues and key shallengev of the secure 5G movile backhaul arcmitecture. The compariskns of the fxisting state-of-the-art solutions for secure mobile backhaul, together with their major contrjbutoons vare yeen ex[lorxd. Furthermore, the paper discussed various key isrues reldtzd to Quality of Service (QoS), tojting and schgduling, resource managemenf, capaciyy enhancement, latency, security-kanagement, and handovers using mwchanjsms lpie Software Defined Metworning qnd miloimeber Wave technulobics. Moreoveg, tht trcils of research challenges and fyture eigections are addmtionally prcstnted.\ \ **Heywords: 5G, Mobils Badnhaul, Security

can be seen in Fig.3 that the decreases the dimension Near the falling the successful output is to that of right successful output, as $N$ grows the proportion occupied by the wrong successful output becomes more significant, as well the noise entropy. This causes a decrease in the transmission. In summary, we proposed experimental to high-dimensional MDI-QKD by exploiting the discrete position basis and momentum basis as two MUBs. It has same level of security as that of polarization-based MDI-QKD. Given a high dimension for this experimental proposal, only quantum secret rate be significantly, but also the quantum-Nit error rate coming from the overlap between diffraction peaks of different basis states in X measurement will substantially. Furthermore, transmission decreases an increasing in this drawback can be overcome by quantum relay. A combination of high-dimensional MDI-QKD and optimization methods serve as a rational approach for improvement of secret key generation rate. This work supported by Young fund of Jiangsu Natural Science Foundation of China (SJ216025), National fund incubation Scientific Research Foundation of University of Posts Telecommunications National Science of China No. 61475075). --- abstract: | The rapid involution of the mobile with incipient data networking capabilities and utilization has exponentially increased traffic Such traffic drains key issues in 5G backhaul Security of mobile backhaul utmost however, limited of which have explored such requirement. This paper discusses the design issues and key mobile backhaul architecture. The comparisons of the existing solutions for secure mobile backhaul, together with major contributions have been explored. Furthermore, the paper discussed various key issues to Quality (QoS), routing and scheduling, resource management, capacity enhancement, security-management, and handovers using like Software Defined Networking and millimeter Wave technologies. Moreover, trails research challenges future directions are presented.\ \ **Keywords: Mobile Backhaul, Security

can be seen in Fig.3 that the tranSmission diStancE deCreAsEs wiTh thE dimension $N$. NeaR The fAlling critical point, the ProbaBiLIty oF ThE wronG succesSFuL OUtpUt Is ComPaRAbLe to tHat Of the riGht successFul OuTput, as $N$ grows LArGer, the propOrtIon occupied bY thE wrong SuCceSSful oUtpUt becOmes moRE signiFicant, as wElL As the nOIse entrOPY. THis cAuses a decrease in tHE tRAnsmission. In suMmary, wE hAVe PROpoSed An experimeNtAl schEMe to impLEmENT HigH-Dimensional MDi-QKD by exploITinG the diScRetE PositiOn basIs ANd mOmentum basiS as tWo MUBs. It hAs the sAMe level OF securiTy as thAt oF poLariZAtIoN-baSeD Two-DImEnsIOnaL MDI-QKD. GIvEn A high EnouGH DIMensIon For tHis exPerimental proPosAl, noT OnlY the qUantuM secReT key rAte can Be enlArGed significantlY, but Also that tHe qUaNtuM-NIt errOR rate cOmiNg fRom the oVerlap bETweEn DIFFrAction peaks of diffeReNT BaSis stateS in X baSIs MeASurement WiLl bE redUCEd subStanTIaLly. FurthErmore, THe TrAnsmissIoN distaNcE deCreAses aS An inCreasiNg in dimeNsion, BUt this drawback CAn be overcome bY ThE MEtHOd of QuaNtum relay. A cOmbiNAtioN of hIGh-DimENsionAl MDI-qKd AnD Existed optimization MeThods wOuld sErve as a rationAl approach FOR Further iMproVEmENt of secret key gEneraTion rate. ThIS work is sUpporTed by YouNg fund of JIANgsu NatuRal sciEncE FoUNDaTion of China (SJ216025), nATionAl Fund incUbaTion proJecT (NY217024), sciEntIfIc ResearcH FoundatIoN oF NAnJinG UnivERsity of POsTs aNd telEcommUNicatiOns (NY215034), natiOnAl nAtuRal ScieNCe fOUndaTiOnS of CHinA (GRant NO. 61475075). --- absTRacT: | The rapId involutIon OF the MoBiLe generAtion with inciPiEnt data netWoRkiNg capaBILities anD utilization has exponentIAlly incReaSed thE datA traffic vOluMes. SucH trAFfic drAins vaRious KeY isSUEs in 5G MOBiLe bAcKhaul netwoRKS. SeCuritY oF mobIle backHaul is of utmost impoRTanCe; however, therE arE a liMITeD nuMBeR Of aRtICleS, WHich have exploreD such a requIrEMeNt. This papeR DisCuSses the PotentiAl desIGn issueS and key chAllenges oF tHe seCURe 5G Mobile backHaul archItecture. THE compARiSons oF thE existInG stAte-of-The-art SOluTions For secUrE mobilE backHaUl, togethEr with their major contribUtions Have bEen Explored. FUrtHErmOre, the papEr diScussed varIouS keY issuEs rELated To QuALiTy oF serviCe (Qos), Routing anD ScHedULInG, resource maNAGEmeNt, capAciTY enhanCemeNt, latency, security-MAnagement, and haNdovERS usIng MEchaNiSms like SoftwarE DeFiNED NetworkInG and millimeTer Wave tEcHNologIes. MorEover, tHe trailS OF rESearch ChalLenGes and futUre DiREctions ArE aDDitionAlly PrEsenteD.\ \ **KeywoRDs: 5G, MOBIle Backhaul, SecurIty

can be seen in Fig.3 that the trans missi ondis ta ncedecr eases with the dime nsion $N$. Near the fa lling c r itic a lpoint , the p r ob a b ili ty o f t he wr ong s ucc essfuloutput iscom pa rable to tha t o f the righ t s uccessful ou tpu t, as$N $ g r ows l arg er, t he pro p ortion occupied b y the w r ong suc c e ss fuloutput becomes mo r es ignificant, as wellas th e noi seentropy. T hi s cau s es a de c re a s e in the transmiss ion. In su m mar y, weha vep ropose d anex p eri mental sche me t o impleme nt hig h -dimens i onal MD I-QKDbyexp loit i ng t hedi s cre t epos i tio n basisan dmomen tumb a s i s as tw o MU Bs. I t has the sam e l evel ofsecur ity a s th at of p olariz ation -b ased two-dimens iona l MDI-QKD . G iv enahighe noughdim ens ion for this e x per im e n t al proposal, not onl yt h equantumsecret ke yr ate canbe en larg e d sign ific a nt ly, butalso t h at t he quan tu m-Niter ror ra te co m ingfrom t he overl ap be t ween diffracti o n peaks of di f fe r e nt basi s s tates in Xbasi s mea sure m en t w i ll be redu ce d s u bstantially. Furthe rm ore, t he tr ansmission di stance dec r e a ses as a n in c re a sing in dimens ion,but this d r awback c an be overcom e by them e thod ofqua ntu m r ela y . A combinationo f hig h- dimensi ona l MDI-Q KDand ex ist ed optimiza tion met ho ds w ou ldserve as a rat io nal a ppr oachf or fur therimpr ov em e ntof secr e tk e y ge ne ra tion ra te . Th is w o rkis supp orted byYou n g fu nd o f Jiang su Natural Sc ie nce Founda ti onof Chi n a (SJ2160 25), National fund incu b ation p roj ect ( NY21 7024), Sc ien tificRes e arch F oundat ion o fNan j i ng Un i v er sit yof Posts a n d Te lecom mu nica tions ( NY215034), Nationa l Na tural Science Fo unda t i on s o f C h ina ( G ran t No. 61475075). --- abstr ac t :| Ther api dinvolut ion ofthe m o bile ge nerationwith inci pi entd a tanetworking capabil ities and utili z at ion h asexpone nt ial ly in crease d th e dat a traf fi c volu mes.Su ch traff ic drains various key i ssuesin 5G mo bile back hau l ne tworks. S ecur ity of mob ile ba ckhau l i s of u tmos t i mpo r tance ; ho w ever, the r eare a l imited numb e r ofartic les , which hav e explored such a requirement. T hisp a per di s cuss es the potential de si g n issuesan d key chall enges of t h e sec ure 5G mobil e backh a u la rchite ctur e.The compa ris on s of the e xi s ting s tate -o f-the- art so l utio n s for secure mobi le ba c k haul, tog ether w ith the i r ma jor contri butions hav e been exp lored . Furth er more,the p aper discu s sed vario us ke y issue srela ted to Qu alit y of Se rvic e(Qo S), routi n g a n dsc h edu ling , res ou rcemanagemen t , capaci tye nhancem en t,l a tency, se c u rity-manag eme nt, a n d handovers usin g m e chani sms li ke Sof tware D e fin ed Networ kin g and milli meter Wav e te ch nolo gies. Mo re over ,the t rails of resear ch chall enges an d future di r ecti o ns are a dditionallyprese nt e d.\ \ **Key w ord s:5G, Mob ile Bac k haul, Security

can_be seen_in Fig.3 that the_transmission distance_decreases_with the_dimension_$N$. Near the_falling critical point,_the probability of the_wrong successful output_is_comparable to that of the right successful output, as $N$ grows larger, the proportion_occupied_by the_wrong_successful_output becomes more significant, as_well as the noise entropy._This causes_a decrease in the transmission. In summary, we have_proposed_an experimental scheme_to implement high-dimensional MDI-QKD by exploiting the discrete position_basis and momentum basis as two_MUBs. It has_the_same_level of security as_that of polarization-based two-dimensional MDI-QKD. Given_a high enough dimension for this_experimental proposal, not only the quantum secret_key rate can be enlarged significantly,_but also that the quantum-Nit_error rate_coming from the overlap between_diffraction peaks of_different basis_states in X_basis measurement will be reduced substantially._Furthermore, the transmission_distance decreases as an increasing in_dimension,_but this drawback_can_be_overcome by_the method of_quantum_relay. A_combination_of high-dimensional MDI-QKD and existed optimization_methods_would serve as a rational approach for_further improvement of secret_key_generation rate. This work is_supported by Young fund of_Jiangsu Natural Science Foundation of China_(SJ216025), National_fund incubation_project (NY217024), Scientific Research Foundation of Nanjing University of Posts and_Telecommunications (NY215034), National Natural Science Foundations_of China (Grant No._61475075). --- abstract:_| _ _The_rapid involution_of the mobile generation with incipient data_networking capabilities_and utilization has exponentially increased the_data traffic volumes. Such_traffic_drains various key issues in 5G_mobile backhaul networks. Security of mobile_backhaul is of utmost importance;_however,_there_are a limited number of_articles, which have explored such a_requirement. This paper_discusses the potential design issues and key_challenges_of the secure 5G mobile backhaul_architecture._The comparisons of the existing state-of-the-art_solutions_for_secure mobile backhaul, together with_their major contributions have been explored._Furthermore, the paper discussed various key issues related to_Quality of Service_(QoS), routing and scheduling, resource_management,_capacity_enhancement, latency, security-management, and handovers using mechanisms like Software Defined_Networking and_millimeter Wave technologies._Moreover, the trails of research challenges and future directions are_additionally presented.\ \ _ **Keywords: 5G, Mobile Backhaul, Security

mag\] [@holland01]. As another example, the expected $B$ band magnitude of the GRB 010921 afterglow would be 3 magnitudes fainter than the host magnitude reported in @price02 at the time of the observations 21 days after the burst. Ideally, the magnitudes of a host should be derived using one consistent photometric technique for all filters. For example, in the case of aperture photometry the magnitudes should be derived using the same aperture. We can not be sure that this is the case for the magnitudes given in Table \[tab:allmag\]. In the cases where the hosts are more extended than point sources the effect should be negligible as long as the authors have applied a large enough aperture for deriving the host magnitudes. Furthermore, one should note that we have restricted ourselves to analysing only the bright end of the luminosity function since it is easier to perform multiband observations of the brightest hosts. Only little is known about the nature of the fainter host galaxies [@berger02b; @jaunsen03; @hjorth03b]. [llll]{} Host & Filter (Mag) & Telescope & Reference\ & $B>26.08$ &VLT &[@sok01]\ & $V=25.77\pm0.2$ &HST/STIS & [@gal00b]\ & $R_c=25.22\pm0.2$ &HST/STIS &\ & $I=24.4\pm0.2$ & HST/WFPC2 &[@fruchter99]\ & $H=23.2\pm$0.3 &HST/NICMOS2\ & $K=22.6\pm$0.2 & NIRC/Keck I &[@chary01]\ & $B=25.89\pm$0.19 & BTA &[@sok01]\ &$V=25.34\pm$0.22 &BTA\ &$R_c=25.06\pm0.17$&BTA\ &$I_c=24.11\pm0.25$&BTA\ &$K=22.7\pm0.2$& Keck I/NIRC& [@chary01]\ & $B=23.40\pm0.12$ &BTA&[@sok01

mag\ ] [ @holland01 ]. As another example, the expected $ B$ band magnitude of the GRB   010921 afterglow would be 3 magnitude fainter than the server magnitude reported in @price02 at the time of the observation 21 days after the burst. Ideally, the magnitude of a host should be derived use one consistent photometric technique for all filter. For example, in the case of aperture photometry the order of magnitude should be derive using the like aperture. We cannot be sure that this is the shell for the magnitude given in Table   \[tab: allmag\ ]. In the cases where the hosts are more extensive than point sources the effect should be negligible equally long as the authors have put on a large enough aperture for deriving the server magnitudes. Furthermore, one should note that we have restricted ourselves to analysing only the bright end of the luminosity function since it is easier to perform multiband observations of the brightest hosts. Only little is known about the nature of the fainter host galaxy [ @berger02b; @jaunsen03; @hjorth03b ]. [ llll ] { } Host & Filter (Mag) & Telescope & Reference\ & $ B>26.08 $ & VLT & [ @sok01]\ & $ V=25.77\pm0.2 $ & HST / STIS & [ @gal00b]\ & $ R_c=25.22\pm0.2 $ & HST / STIS & \ & $ I=24.4\pm0.2 $ & HST / WFPC2 & [ @fruchter99]\ & $ H=23.2\pm$0.3 & HST / NICMOS2\ & $ K=22.6\pm$0.2 & NIRC / Keck I & [ @chary01]\ & $ B=25.89\pm$0.19 & BTA & [ @sok01]\ & $ V=25.34\pm$0.22 & BTA\ & $ R_c=25.06\pm0.17$&BTA\ & $ I_c=24.11\pm0.25$&BTA\ & $ K=22.7\pm0.2 $ & Keck I / NIRC & [ @chary01]\ & $ B=23.40\pm0.12 $ & BTA&[@sok01

mag\] [@holland01]. As another examkle, the expected $Y$ band kagnithde of tfe GRB 010921 afterglow would be 3 megnirudes fainter than the host magnitudv reportee in @price02 at tis time of the lbsexvetions 21 days afjer the bursd. Ideally, the mdgvicudes of a host should be derived usyng one clnsistent photjmetgis tedhnique for all filters. For exampls, in tht case of aperture photometry the magnitudes shokld be derived usijg the same apewrure. We can vot be surt chat this ia the case for the magnitudes gkven nn Table \[tab:qlomah\]. In the casxs whege the hosts are more extendrd than point xoucces the effect should be negligible as lond as the duchors have applied a oaege etougv apdetufe roc dsrivinh tie host maghitudes. Furtyermore, one should moev that we havs restwisted ourselves to analysing only the brpght end of the luminosity dunction since it is gasier to [erform multiband observations of the brightest hmsts. Kvly llttld id known about the nature of the fainter host dzlsxpes [@berger02b; @jaunscn03; @hjorth03b]. [llll]{} Hosy & Fojter (Mag) & Telgscope & Rererence\ & $B>26.08$ &VLT &[@sok01]\ & $G=25.77\pm0.2$ &HST/FTIS & [@gal00b]\ & $R_c=25.22\pi0.2$ &HSY/STIS &\ & $I=24.4\pm0.2$ & HST/WFPC2 &[@fruchtee99]\ & $H=23.2\pm$0.3 &HST/NICIIS2\ & $K=22.6\pm$0.2 & NIRC/Keck I &[@chary01]\ & $B=25.89\pm$0.19 & BTA &[@sok01]\ &$V=25.34\lm$0.22 &BTA\ &$R_c=25.06\pm0.17$&BTA\ &$I_c=24.11\pm0.25$&BTA\ &$K=22.7\pm0.2$& Keci I/NIRC& [@chagy01]\ & $B=23.40\pm0.12$ &BTZ&[@rok01

mag\] [@holland01]. As another example, the expected magnitude the GRB afterglow would be host reported in @price02 the time of observations 21 days after the burst. the magnitudes of a host should be derived using one consistent photometric technique all filters. For example, in the case of aperture photometry the magnitudes should derived the aperture. can not be sure that this is the case for the magnitudes given in Table \[tab:allmag\]. the cases where the hosts are more extended point sources the effect be negligible as long as authors applied a enough for the host magnitudes. one should note that we have restricted ourselves to analysing only the bright end of the luminosity since it to perform observations the hosts. Only little about the nature of the fainter @jaunsen03; @hjorth03b]. [llll]{} Host & Filter (Mag) & & Reference\ $B>26.08$ &VLT &[@sok01]\ & $V=25.77\pm0.2$ &HST/STIS [@gal00b]\ & $R_c=25.22\pm0.2$ &HST/STIS &\ & $I=24.4\pm0.2$ & &[@fruchter99]\ & $H=23.2\pm$0.3 &HST/NICMOS2\ & $K=22.6\pm$0.2 & NIRC/Keck I &[@chary01]\ & $B=25.89\pm$0.19 & BTA &[@sok01]\ &$R_c=25.06\pm0.17$&BTA\ &$I_c=24.11\pm0.25$&BTA\ &$K=22.7\pm0.2$& Keck [@chary01]\ & $B=23.40\pm0.12$

mag\] [@holland01]. As another examplE, the expectEd $B$ baNd mAgnItUde oF the gRB 010921 afterglow woULd be 3 Magnitudes fainter than tHe hosT mAGnitUDe ReporTed in @prICe02 AT The TiMe Of tHe OBsErvatIonS 21 days afTer the bursT. IdEaLly, the magnitUDeS of a host shOulD be derived usIng One conSiSteNT photOmeTric tEchniqUE for alL filters. FOr EXample, IN the casE OF aPertUre photometry the mAGnITudes should be dErived UsINg THE saMe aPerture. We cAn Not be SUre that THiS IS The CAse for the magnItudes given IN TaBle \[tab:AlLmaG\]. in the cAses wHeRE thE hosts are moRe exTended thaN point SOurces tHE effect Should Be nEglIgibLE aS lOng As THe aUThOrs HAve Applied a LaRgE enouGh apERTURe foR deRiviNg the Host magnitudeS. FuRtheRMorE, one sHould Note ThAt we hAve resTrictEd Ourselves to analYsinG only the bRigHt End Of The luMInositY fuNctIon sincE it is eaSIer To PERFoRm multiband observaTiONS oF the brigHtest hOStS. ONLy little Is KnoWn abOUT the nAturE Of The faintEr host GAlAxIes [@bergEr02B; @jaunsEn03; @HjoRth03B]. [llll]{} hOst & FIlter (MAg) & TelescOpe & ReFErence\ & $B>26.08$ &VLT &[@sok01]\ & $V=25.77\PM0.2$ &HST/STIS & [@gal00b]\ & $R_C=25.22\Pm0.2$ &hst/StiS &\ & $I=24.4\pM0.2$ & HSt/WFPC2 &[@fruchtEr99]\ & $H=23.2\pM$0.3 &hST/NiCMOs2\ & $k=22.6\pM$0.2 & NIrc/Keck i &[@charY01]\ & $B=25.89\PM$0.19 & Bta &[@sok01]\ &$V=25.34\pm$0.22 &BTA\ &$R_c=25.06\pm0.17$&BTA\ &$I_c=24.11\Pm0.25$&bTA\ &$K=22.7\pm0.2$& keck I/nIRC& [@chary01]\ & $B=23.40\pm0.12$ &BtA&[@sok01

mag\] [@holland01]. As ano ther examp le, t heexp ec ted$B$band magnitude of t he GRB 010921 afterglo w wou ld be 3 ma gnitu des fai n te r tha nth e h os t m agnit ude report ed in @pri ce0 2at the timeo fthe observ ati ons 21 daysaft er the b urs t . Id eal ly, t he mag n itudes of a hos ts houldb e deriv e d u sing one consistent p h ot o metric techniq ue for a l lf i lte rs. For examp le , int he case of a p ert u re photometry the magnit u des shoul dbed erived usin gt hesame apertu re.We can no t be s u re that this is the c ase fo r th e m ag nit ud e s g i ve n i n Ta ble \[ta b: al lmag\ ]. I n t h e ca ses whe re th e hosts are m ore ext e nde d tha n poi nt s ou rcesthe ef fectsh ould be negligi bleas long a s t he au th ors h a ve app lie d a largeenougha per tu r e fo r deriving the hos tm a gn itudes. Furth e rm or e , one sh ou ldnote t hat w e ha v erestrict ed our s el ve s to an al ysingon lythe brig h t en d of t he lumin osity function since it is easiert op e rf o rm m ult iband obser vati o ns o f th e b rig h testhosts .O nl y little is known ab ou t thenatur e of the fain ter host g a l a xies [@b erge r 02 b ; @jaunsen03;@hjor th03b]. [ l lll]{} H ost & Filter(Mag) & T e l escope & Re fer enc e\& $B >26.08$ &VLT& [ @sok 01 ]\ & $V =25 .77\pm0 .2$ &H ST/ STI S& [@gal00 b]\ & $R _c =2 5. 22 \pm 0.2$& HST/STIS & \ & $ I=2 4.4\p m 0.2$ & HST/ WFPC 2&[ @ fru chter99 ] \& $H=2 3. 2\ pm$0 .3&H ST/NI CMOS 2 \ & $K=22. 6\pm$0.2& N I RC/K ec kI &[@ch ary01]\ & $B= 25 .89\pm$0.1 9& B TA &[@ s o k01]\ &$ V=25.34\pm$0.22 &BTA\ & $ R_c=25. 06\ pm0.1 7$&B TA\ &$I_c =24 .11\pm 0.2 5 $&BTA\ &$K=2 2.7\p m0 .2$ & KeckI / NI RC& [ @chary01]\ & $B =23.4 0\ pm0. 12$ &BT A&[@sok01

mag\] [@holland01]._As another_example, the expected $B$_band magnitude_of_the GRB 010921_afterglow_would be 3_magnitudes fainter than_the host magnitude reported_in @price02 at_the_time of the observations 21 days after the burst. Ideally, the magnitudes of a host_should_be derived_using_one_consistent photometric technique for all_filters. For example, in the_case of_aperture photometry the magnitudes should be derived using_the_same aperture. We_can not be sure that this is the case_for the magnitudes given in Table \[tab:allmag\]._In the cases_where_the_hosts are more extended_than point sources the effect should_be negligible as long as the_authors have applied a large enough aperture_for deriving the host magnitudes. Furthermore, one_should note that we have_restricted ourselves_to analysing only the bright_end of the_luminosity function_since it is_easier to perform multiband observations of_the brightest hosts._Only little is known about the_nature_of the fainter_host_galaxies_[@berger02b; @jaunsen03;_@hjorth03b]. [llll]{} Host &_Filter_(Mag) &_Telescope_& Reference\ & $B>26.08$ &VLT &[@sok01]\ & $V=25.77\pm0.2$_&HST/STIS_& [@gal00b]\ & $R_c=25.22\pm0.2$ &HST/STIS &\ & $I=24.4\pm0.2$ &_HST/WFPC2 &[@fruchter99]\ & $H=23.2\pm$0.3 &HST/NICMOS2\ &_$K=22.6\pm$0.2_& NIRC/Keck I &[@chary01]\ &_$B=25.89\pm$0.19 & BTA &[@sok01]\ &$V=25.34\pm$0.22 &BTA\ &$R_c=25.06\pm0.17$&BTA\ &$I_c=24.11\pm0.25$&BTA\ &$K=22.7\pm0.2$&_Keck I/NIRC& [@chary01]\ & $B=23.40\pm0.12$ &BTA&[@sok01

its detection in km$^{2}$ surface area high energy neutrino telescopes. For $2\times 10^{6}\leq E$/GeV $\leq 2\times 10^{7}$, the down ward going high energy cosmic tau neutrinos may produce a double shower signature because of charged current deep inelastic scattering followed by a subsequent hadronic decay of associated tau lepton The double shower event rate for intrinsic (no oscillations/conversions) high energy tau neutrinos originating from GRBs turns out to be small as compared to that due to precession/conversion effects up to a factor of $\sim \, 10^{-5}$. Thus, the high energy neutrino telescopes may possibly provide useful upper bounds on intrinsic properties of neutrinos such as mass, mixing and magnetic moment, etc.. The relevant tau neutrino energy range for detection in km$^{2}$ surface area under water/ice neutrino telescopes may be $2\times 10^{6} {\buildrel < \over {_{\sim}}}\, E/\mbox {GeV}\, {\buildrel < \over {_{\sim}}} \, 2\times 10^{7}$ through characteristic contained but separable double shower events. Observationally, the high energy $\nu_{\tau}$ burst from a GRB may possibly be [*correlated*]{} to the corresponding gamma-ray burst/highest energy cosmic rays (if both have common origin) in time and in direction thus raising the possibility of its detection. If the range of neutrino mixing parameters pointed out in this study is realized terrestially/extraterrestially then a relatively large (energy dependent) $\nu_{\tau}$ flux from GRBs is expected as compared to no oscillation/conversion scenario. #### Acknowledgments. {#acknowledgments..unnumbered} The author thanks Japan Society for the Promotion of Science for financial support. E. Waxman and J. Bahcall, Phys. Rev. Lett. 78 (1997) 2292 ; Phys. Rev. D 59 (1999) 023002. For a recent review, see, E. Waxman, astro-ph/9911395. For a recent review, see, for instance, T. Piran, Nucl. Phys. B (Proc. Suppl.) 70 (1999) 431 and references cited therein. For a latest discussion, see, P. Meszaros, astro-ph/9904038; T. Piran, astro-ph/9907392. B. Paczyński and G. Xu, Ap.

its detection in km$^{2}$ surface area high department of energy neutrino telescope. For $ 2\times 10^{6}\leq E$/GeV $ \leq 2\times 10^{7}$, the gloomy ward go gamey energy cosmic tau neutrinos may grow a bivalent shower signature because of charged current deep inelastic scattering keep up by a subsequent hadronic decay of associated tau lepton The double exhibitor event rate for intrinsic (no oscillations / conversions) high energy tau neutrinos originate from GRBs turns out to be small as compared to that due to precession / conversion effect up to a factor of $ \sim \, 10^{-5}$. Thus, the high energy neutrino telescopes may possibly supply useful upper bounds on intrinsic properties of neutrinos such as mass, mix and magnetic moment, etc.. The relevant tau neutrino energy range for detection in km$^{2}$ surface sphere under body of water / ice neutrino telescopes may be $ 2\times 10^{6 } { \buildrel < \over { _ { \sim}}}\, E/\mbox { GeV}\, { \buildrel < \over { _ { \sim } } } \, 2\times 10^{7}$ through characteristic contained but separable double shower events. Observationally, the high energy $ \nu_{\tau}$ explosion from a GRB may possibly be [ * correlated * ] { } to the corresponding da gamma - re outburst / highest energy cosmic rays (if both have common origin) in time and in focus thus raising the possibility of its detection. If the image of neutrino mixing parameters pointed out in this study is understand terrestially / extraterrestially then a relatively bombastic (energy dependent) $ \nu_{\tau}$ flux from GRBs is expected as compared to no cycle / conversion scenario. # # # # Acknowledgments. { # acknowledgments.. unnumbered } The author thanks Japan Society for the Promotion of Science for financial support. E. Waxman and J. Bahcall, Phys. Rev. Lett. 78 (1997) 2292; Phys. Rev. D 59 (1999) 023002. For a recent review, see, E. Waxman, astro - ph/9911395. For a late review, see, for instance, T. Piran, Nucl. Phys. B (Proc. Suppl .) 70 (1999) 431 and references cited therein. For a up-to-the-minute discussion, see, P. Meszaros, astro - ph/9904038; T. Piran, astro - ph/9907392. B. Paczyński and G. Xu, Ap.

itd detection in km$^{2}$ surfact area high energi beutriio teleacopes. Fur $2\times 10^{6}\leq E$/GeV $\leq 2\times 10^{7}$, tye doqn ward going high enefgy cosmib tau neurriniw may prodnde a domyle sglwer wignature becaose of chargad current dee[ knzlastic scattering followed by a subfequent hwdronic decay jf axfocizneb tau lepton The double shower ebent raue for intrinsic (np oscillations/conversions) jigh energy tau neutrijos originajjng drom GRBs tufns out to be small as compared to that due to precesrion/cpnversion gydechv up to a fector jf $\sim \, 10^{-5}$. Thus, the hich enerby neutrino tekesropew may possibly providx useful upper boundf on intrhnaic properties of nwutritos vuch qs oasa, kising ajd jagnetic mkment, etc.. Tye relevant tau neuurigi energy rangs for qeeection in km$^{2}$ surface area under water/ibe nsutrino telescopes may ve $2\times 10^{6} {\buildrel < \oger {_{\sim}}}\, E/\ibox {GeV}\, {\buildrel < \over {_{\sim}}} \, 2\times 10^{7}$ through characderisviz cinbainde hut separable double shower events. Observationwmlu, nhe high energy $\nm_{\tau}$ burst from a BRH kwy possibly bg [*correlatsd*]{} to the correspojding gwmma-rqy burst/hyghext energy cosmic rays (if borh have commjb origin) in time aud in directnon thos raixing the possibility of its detection. Lf the rahee of neutrino mkxikg [arameters pointed out in ehis studb is xealized tertestialjy/extratergestially then a relativfly lcrge (anergy depfndent) $\nu_{\tau}$ flux from GRBs is xepected as cokpdreg to no jscilkation/conversyon scenario. #### Aeknowledyments. {#acknowledfments..uinumbered} The author thankv Japan Socievy for thq Primotuon of Rzience for finsncial suipjrt. E. Waxman and J. Bahcell, Khgs. Rev. Lett. 78 (1997) 2292 ; Phiw. Rev. D 59 (1999) 023002. For a fecqnn rxview, vee, E. Waxman, astfo-pf/9911395. For x recenu rcvidw, sre, for instance, T. Pisan, Hucl. Phys. B (Proc. Sipil.) 70 (1999) 431 and referenses cited thetein. For a latest fiscuvsiun, ser, P. Meszaros, astro-ph/9904038; T. Piran, astrk-ph/9907392. B. Pacxyńshi and G. Xu, Wp.

its detection in km$^{2}$ surface area high telescopes. $2\times 10^{6}\leq $\leq 2\times 10^{7}$, energy tau neutrinos may a double shower because of charged current deep inelastic followed by a subsequent hadronic decay of associated tau lepton The double shower rate for intrinsic (no oscillations/conversions) high energy tau neutrinos originating from GRBs turns to small compared that due to precession/conversion effects up to a factor of $\sim \, 10^{-5}$. Thus, the high neutrino telescopes may possibly provide useful upper bounds intrinsic properties of neutrinos as mass, mixing and magnetic etc.. relevant tau energy for in km$^{2}$ surface under water/ice neutrino telescopes may be $2\times 10^{6} {\buildrel < \over {_{\sim}}}\, E/\mbox {GeV}\, {\buildrel < \over \, 2\times characteristic contained separable shower Observationally, the high burst from a GRB may possibly the corresponding gamma-ray burst/highest energy cosmic rays (if have common in time and in direction thus the possibility of its detection. If the range neutrino mixing parameters pointed out in this study is realized terrestially/extraterrestially then a relatively large $\nu_{\tau}$ flux from GRBs expected as compared no scenario. Acknowledgments. The author Japan Society for the Promotion of Science for financial support. E. and J. Bahcall, Phys. Rev. Lett. 78 (1997) 2292 ; D (1999) 023002. For recent review, see, E. astro-ph/9911395. a recent review, see, T. Nucl. Suppl.) (1999) and references cited therein. a latest discussion, see, P. astro-ph/9904038; T. Piran, astro-ph/9907392. Ap.

its detection in km$^{2}$ surface arEa high enerGy neuTriNo tElEscoPes. FOr $2\times 10^{6}\leq E$/GeV $\LEq 2\tiMes 10^{7}$, the down ward going higH enerGy COsmiC TaU neutRinos maY PrODUce A dOuBle ShOWeR signAtuRe becauSe of chargeD cuRrEnt deep inelaSTiC scatterinG foLlowed by a subSeqUent haDrOniC Decay Of aSsociAted taU Lepton the double ShOWer eveNT rate foR INtRinsIc (no oscillations/cONvERsions) high enerGy tau nEuTRiNOS orIgiNating from gRbs turNS out to bE SmALL As cOMpared to that dUe to precessIOn/cOnversIoN efFEcts up To a faCtOR of $\Sim \, 10^{-5}$. Thus, the hIgh eNergy neutRino teLEscopes MAy possiBly proVidE usEful UPpEr BouNdS On iNTrInsIC prOperties Of NeUtrinOs suCH AS Mass, MixIng aNd magNetic moment, etC.. ThE relEVanT tau nEutriNo enErGy ranGe for dEtectIoN in km$^{2}$ surface areA undEr water/icE neUtRinO tElescOPes may Be $2\tImeS 10^{6} {\buildrEl < \over {_{\sIM}}}\, E/\mBoX {gEv}\, {\bUildrel < \over {_{\sim}}} \, 2\timeS 10^{7}$ tHROuGh characTeristIC cOnTAined but SeParAble DOUble sHoweR EvEnts. ObseRvatioNAlLy, The high EnErgy $\nu_{\TaU}$ buRst From a grB maY possiBly be [*corRelatED*]{} to the correspoNDing gamma-ray bURsT/HIgHEst eNerGy cosmic rayS (if bOTh haVe coMMoN orIGin) in Time aNd IN dIRection thus raising tHe PossibIlity Of its detectioN. If the rangE OF Neutrino MixiNG pARameters pointeD out iN this study IS realizeD terrEstially/ExtraterrESTially thEn a RelAtiVelY LArGe (energy depenDENt) $\nu_{\TaU}$ flux frOm GrBs is exPecTed As cOmpArEd to no oscIllation/CoNvErSiOn sCenarIO. #### AcknowlEdGmeNtS. {#acKnowlEDgmentS..unnuMberEd} thE AutHor thanKS JAPAn SoCiEtY for The prOmotiOn of sCieNce for fInancial sUppORt. E. WAxMaN and J. BaHcall, Phys. Rev. LEtT. 78 (1997) 2292 ; Phys. Rev. D 59 (1999) 023002. FOr A reCent reVIEw, see, E. WaXman, astro-ph/9911395. For a recent reVIew, see, fOr iNstanCe, T. PIran, Nucl. PHys. b (Proc. SUppL.) 70 (1999) 431 And refErenceS citeD tHerEIN. For a LATeSt dIsCussion, see, p. mEszAros, aStRo-ph/9904038; t. Piran, aStro-ph/9907392. B. Paczyński anD g. Xu, ap.

its detection in km$^{2}$ surface a rea h igh en er gy n eutr ino telescopes . For $2\times 10^{6}\leq E $/GeV $ \ leq2 \t imes10^{7}$ , t h e do wn w ard g o in g hig h e nergy c osmic tauneu tr inos may pro d uc e a double sh ower signatu rebecaus eofc harge d c urren t deep inelas tic scatt er i ng fol l owed by a s ubse quent hadronic de c ay of associatedtau le pt o nT hedou ble shower e ventr ate for in t r i nsi c (no oscillat ions/conver s ion s) hig hene r gy tau neut ri n osoriginating fro m GRBs tu rns ou t to bes mall as compa red to tha t d ue to p r ece s si on/ c onv ersion e ff ec ts up toa f a ctor of $\s im \, 10^{-5}$. Th us, the hig h ene rgy n eutr in o tel escope s may p ossibly provide use ful upper bo un dson intr i nsic p rop ert ies ofneutrin o s s uc h a smass, mixing and m ag n e ti c moment , etc. . T he relevant t auneut r i no en ergy ra nge fordetect i on i n km$^{ 2} $ surf ac e a rea unde r wat er/ice neutrin o tel e scopes may be$ 2\times 10^{6 } { \ bu i ldre l < \over {_{\ sim} } }\,E/\m b ox {G e V}\,{\bui ld r el < \over {_{\sim}}}\, 2\tim es 10 ^{7}$ through character i s t ic conta ined bu t separable dou ble s hower even t s. Obse rvati onally,the highe n ergy $\n u_{ \ta u}$ bu r s tfrom a GRB ma y poss ib ly be [ *co rrelate d*] {}tothe c orrespond ing gamm a- ra ybu rst /high e st energ ycos mi c r ays ( i f both have com mo no rig in) int im e andin d irec tio nthusrais i ngthe pos sibilityofi ts d et ec tion. I f the range o fneutrino m ix ing param e t ers poin ted out in this study i s realiz edterre stia lly/extra ter restia lly then a relat ively l arg e (ener g y d epe nd ent) $\nu_ { \ tau }$ fl ux fro m GRBsis expected as com p are d to no oscil lat ion/ c o nv ers i on sce na r io. #### Acknowledg ments. {#a ck n ow ledgments. . unn um bered} The au thort hanks J apan Soci ety for t he Pro m o tio n of Scien ce for f inancials uppor t . E. W axm an and J . B ahcal l, Phy s . R ev. L ett. 7 8(1997) 2292 ; Phys. R ev. D 59 (1999) 023002. For a rece ntreview, s ee, E.Waxman, a stro -ph/991139 5.For a re cen t revi ew,s ee , f o r ins tanc e , T. Pira n ,Nuc l . P hys. B (Pro c . Sup pl.)70( 1999)431and references ci t ed therein. Fo r al a tes t d i scus si on, see, P. Me sza ro s , astro-p h/ 9904038; T. Piran,as t ro-ph /99073 92. B. Paczyń s k ia nd G.Xu,Ap.

its_detection in_km$^{2}$ surface area high_energy neutrino_telescopes._For $2\times_10^{6}\leq_E$/GeV $\leq 2\times_10^{7}$, the down_ward going high energy_cosmic tau neutrinos_may_produce a double shower signature because of charged current deep inelastic scattering followed by_a_subsequent hadronic_decay_of_associated tau lepton The double shower_event rate for intrinsic (no_oscillations/conversions) high_energy tau neutrinos originating from GRBs turns out_to_be small as_compared to that due to precession/conversion effects up to_a factor of $\sim \, 10^{-5}$._Thus, the high_energy_neutrino_telescopes may possibly provide_useful upper bounds on intrinsic properties_of neutrinos such as mass, mixing_and magnetic moment, etc.. The relevant tau_neutrino energy range for detection in_km$^{2}$ surface area under water/ice_neutrino telescopes_may be $2\times 10^{6} {\buildrel_< \over {_{\sim}}}\,_E/\mbox {GeV}\, {\buildrel_< \over {_{\sim}}}_\, 2\times 10^{7}$ through characteristic contained_but separable double_shower events. Observationally, the high energy $\nu_{\tau}$_burst_from a GRB_may_possibly_be [*correlated*]{}_to the corresponding_gamma-ray_burst/highest energy_cosmic_rays (if both have common origin)_in_time and in direction thus raising the_possibility of its detection._If_the range of neutrino_mixing parameters pointed out in_this study is realized terrestially/extraterrestially then_a relatively_large (energy_dependent) $\nu_{\tau}$ flux from GRBs is expected as compared to no_oscillation/conversion scenario. #### Acknowledgments. {#acknowledgments..unnumbered} The author thanks_Japan Society for the_Promotion of_Science_for financial support. E._Waxman_and J._Bahcall, Phys. Rev. Lett. 78 (1997) 2292_; Phys._Rev. D 59 (1999) 023002. For_a recent review, see,_E._Waxman, astro-ph/9911395. For a recent review,_see, for instance, T. Piran, Nucl._Phys. B (Proc. Suppl.) 70_(1999)_431_and references cited therein. For_a latest discussion, see, P. Meszaros,_astro-ph/9904038; T. Piran,_astro-ph/9907392. B. Paczyński and G. Xu, Ap.

, Phys. Rev. [**D67**]{}, 094002 (2003), Phys. Rev. [**D69**]{}, 074026 (2004) and Phys. Rev. [**D74**]{} 114007 (2006). X. Artru, hep-ph/0207309 (2002). P. Abbon [*et al.*]{} \[COMPASS collaboration\] hep-ex/0703049, accepted by NIM [**A**]{}. --- author: - 'Laurent Chevillard, Raoul Robert and Vincent Vargas' title: A Stochastic Representation of the Local Structure of Turbulence --- Introduction ============ Fully developed turbulent flows are omnipresent in Nature ([e.g.]{} meteorology) and engineering ([e.g.]{} combustion). Despite an apparent complexity, it turns out that these flows exhibit universal statistical properties such as the Kolmogorov $k^{-5/3}$-law observed on the power spectrum [@K41] and the intermittent nature of longitudinal and transverse velocity fluctuations [@Fri95]. Modern [developments]{} of experimental and numerical facilities [@Tsino01; @Wal09] have furthermore underlined the peculiar and universal geometry of turbulence. From a mathematical viewpoint, our knowledge of Navier-Stokes and Euler equations is still poor and this makes the situation far from clear if we look for a relevant model of turbulent flows. [Indeed, one can prove the existence of dissipative solutions of Euler equations and the associated inertial dissipation process has been studied [@DucRob00]. As a consequence the introduction of viscosity is not the only way we have to build realistic model of turbulent flows.]{} Some progresses have been achieved recently in the understanding of the universality of the small scales of turbulence [by]{} studying the Lagrangian dynamics of the velocity gradient tensor $A_{ij} = \partial_ju_i$, where $\textbf{u}$ denotes the velocity (see [@CheA2] and references therein). While this approach takes completely into account the local interactions governing the dynamics of $\textbf{A}$, it requires closures for both the pressure Hessian and viscous term. As far as we know, provided closures miss at least partially the nonlocal nature of pressure [@CheA2]. An alternative approach would be devoted to consider the spatial distribution of the vectorial velocity field. Very few theoretical works have focused on this difficult, although important, aspect. Nonetheless, it has been shown that taking into account numerically the short time advection of fluid particles for all

, Phys. Rev. [ * * D67 * * ] { }, 094002 (2003), Phys. Rev. [ * * D69 * * ] { }, 074026 (2004) and Phys. Rev. [ * * D74 * * ] { } 114007 (2006). X. Artru, hep - ph/0207309 (2002). P. Abbon [ * et   al. * ] { } \[COMPASS collaboration\ ] hep - ex/0703049, accepted by NIM [ * * A * * ] { }. --- author: -' Laurent Chevillard, Raoul Robert and Vincent Vargas' title: A Stochastic Representation of the Local Structure of Turbulence --- initiation = = = = = = = = = = = = amply developed turbulent flows are omnipresent in Nature ([ e.g. ] { } meteorology) and technology ([ e.g. ] { } combustion). Despite an apparent complexity, it turns out that these flows expose universal statistical properties such as the Kolmogorov $ k^{-5/3}$-law observe on the power spectrum [ @K41 ] and the intermittent nature of longitudinal and transverse velocity fluctuation [ @Fri95 ]. Modern [ developments ] { } of experimental and numerical facility [ @Tsino01; @Wal09 ] have furthermore underlined the curious and universal geometry of turbulence. From a mathematical point of view, our knowledge of Navier - Stokes and Euler equations is still inadequate and this makes the situation far from clear if we look for a relevant exemplar of turbulent flows. [ Indeed, one can prove the existence of dissipative solutions of Euler equation and the associated inertial dissipation process has been studied [ @DucRob00 ]. As a consequence the introduction of viscosity is not the only way we have to build realistic model of turbulent flows. ] { } Some progresses have been achieved recently in the understanding of the universality of the small plate of turbulence [ by ] { } studying the Lagrangian dynamics of the speed gradient tensor $ A_{ij } = \partial_ju_i$, where $ \textbf{u}$ announce the speed (see [ @CheA2 ] and references therein). While this approach takes completely into account the local interactions governing the moral force of $ \textbf{A}$, it requires closures for both the pressure Hessian and viscous term. As far as we know, provided closures miss at least partially the nonlocal nature of pressure [ @CheA2 ]. An alternative overture would be devoted to view the spatial distribution of the vectorial velocity field. Very few theoretical works have focused on this difficult, although authoritative, aspect. Nonetheless, it has been shown that assume into account numerically the short time advection of fluent particles for all

, Phjs. Rev. [**D67**]{}, 094002 (2003), Phys. Rev. [**D69**]{}, 074026 (2004) xnd Phys. Rev. [**D74**]{} 114007 (2006). X. Artcu, hep-pg/0207309 (2002). P. Abbun [*et al.*]{} \[COMPASS collaboratioi\] hep-ex/0703049, axcepted by NIM [**A**]{}. --- authof: - 'Laurent Chevillqrd, Caoul Robert and Vincent Vargaa' bitle: E Stochastic Rekresentation of the Local Vtfueture of Turbulence --- Introduction ============ Fully develolef turbulent fljws swe ojnipresent in Nature ([e.g.]{} meteorologg) and eigineering ([e.g.]{} cpmbustion). Despite an apparfnt fomplexity, it turnd out that jgesq flows exhibkt universal statisticzl properties such as the Kolmoeorov $k^{-5/3}$-law obsetrwd lt the power specnrum [@K41] and thc interkittent nature of lonnitudmnal and transverse velocmty fluctuations [@Fri95]. Modern [davzlopments]{} of experimebtql ang nukerizql wacjlmtiss [@Tsijo01; @Xal09] have fudthermore ubderlined the peculoaw and universam geomqtwy of turbulence. From a mathematical viefpojnt, our knowledge of Nacier-Stokes and Euler gquations ys still poor and this makes the situation far frmm clxaf iy we luik for a relevant model of turbulent flows. [Indeqs, pnv can prove the ewistence of dissipstlvr solutions of Euler eqhations and the asdociateq ineetial disfipayion process has been studiwd [@DucRob00]. As q consequence the nntroduction of fiscoxity is not the only wav we hzve to builf realistjz model of turbuuenn flmws.]{} Some progresses have bqen achieted rzcently kn tne undqrstanding of tma universality of hhe slanl scales lf turbulence [by]{} studying the Lejrangian dynakiws mf the vzlocitn gradient tensjr $A_{ij} = \partiak_ju_i$, whzre $\tebtbf{u}$ denones the vxlocity (see [@SheA2] and refesgnces therein). While tris qppriach tayds completely onto accolnu the locao interactions govcrnine the dynamics oy $\ueztbf{A}$, it requites cljslrex smr both the [resrurd Hesskan and visgour tetm. As far as we know, probided closures misx wt least partialjy the nonlocsl nature of presskre [@CieA2]. An altetnajive approach would be devoted fo considfr bhe spatial dystrlbutyon of the vectorial velocity field. Very few theocetical works have focosed on this difficulj, akthough impprtane, aspect. Tonetheless, it has bwen shown that taling into account numsricalny thf short time advection of fluid particles for all

, Phys. Rev. [**D67**]{}, 094002 (2003), Phys. 074026 and Phys. [**D74**]{} 114007 (2006). Abbon al.*]{} \[COMPASS collaboration\] accepted by NIM --- author: - 'Laurent Chevillard, Raoul and Vincent Vargas' title: A Stochastic Representation of the Local Structure of Turbulence Introduction ============ Fully developed turbulent flows are omnipresent in Nature ([e.g.]{} meteorology) and ([e.g.]{} Despite apparent it turns out that these flows exhibit universal statistical properties such as the Kolmogorov $k^{-5/3}$-law observed the power spectrum [@K41] and the intermittent nature longitudinal and transverse velocity [@Fri95]. Modern [developments]{} of experimental numerical [@Tsino01; @Wal09] furthermore the and universal geometry turbulence. From a mathematical viewpoint, our knowledge of Navier-Stokes and Euler equations is still poor and this the situation clear if look a model of turbulent one can prove the existence of Euler equations and the associated inertial dissipation process been studied As a consequence the introduction of is not the only way we have to realistic model of turbulent flows.]{} Some progresses have been achieved recently in the understanding of of the small scales turbulence [by]{} studying Lagrangian of velocity tensor $A_{ij} \partial_ju_i$, where $\textbf{u}$ denotes the velocity (see [@CheA2] and references therein). this approach takes completely into account the local interactions governing of it requires closures both the pressure Hessian viscous As far as we closures at nonlocal of [@CheA2]. An alternative approach be devoted to consider the distribution of the vectorial works have focused on this difficult, although important, Nonetheless, it has been shown that taking account numerically the short time advection of fluid particles for all

, Phys. Rev. [**D67**]{}, 094002 (2003), Phys. Rev. [**D69**]{}, 074026 (2004) and Phys. ReV. [**D74**]{} 114007 (2006). X. Artru, heP-ph/0207309 (2002). P. ABboN [*et Al.*]{} \[cOMPaSS cOllaboration\] heP-Ex/0703049, acCepted by NIM [**A**]{}. --- author: - 'LaurEnt ChEvILlarD, raOul RoBert and vInCENt VArGaS' tiTlE: a STochaStiC RepresEntation of The loCal Structure OF TUrbulence --- INtrOduction ============ FullY deVelopeD tUrbULent fLowS are oMnipreSEnt in NAture ([e.g.]{} meTeORology) ANd enginEERiNg ([e.g.]{} Combustion). Despite AN aPParent complexiTy, it tuRnS OuT THat TheSe flows exhIbIt uniVErsal stATiSTICal PRoperties such As the KolmogORov $K^{-5/3}$-law obSeRveD On the pOwer sPeCTruM [@K41] and the intErmiTtent natuRe of loNGitudinAL and traNsversE veLocIty fLUcTuAtiOnS [@fri95]. mOdErn [DEveLopments]{} Of ExPerimEntaL AND NumeRicAl faCilitIes [@Tsino01; @Wal09] haVe fUrthERmoRe undErlinEd thE pEculiAr and uNiverSaL geometry of turbUlenCe. From a maTheMaTicAl ViewpOInt, our KnoWleDge of NaVier-StoKEs aNd eULEr Equations is still poOr AND tHis makes The sitUAtIoN Far from cLeAr iF we lOOK for a ReleVAnT model of TurbulENt FlOws. [IndeEd, One can PrOve The ExistENce oF dissiPative soLutioNS of Euler equatiONs and the assocIAtED InERtiaL diSsipation prOcesS Has bEen sTUdIed [@dUcRob00]. as a coNsEQuENce the introduction oF vIscosiTy is nOt the only way wE have to buiLD REalistic ModeL Of TUrbulent flows.]{} SOme prOgresses haVE been achIeved Recently In the undeRSTanding oF thE unIveRsaLITy Of the small scaLES of tUrBulence [By]{} sTudying The lagRanGiaN dYnamics of The velocItY gRaDiEnt TensoR $a_{ij} = \partiAl_Ju_i$, WhEre $\TextbF{U}$ denotEs the VeloCiTy (SEe [@CHeA2] and rEFeRENces ThErEin). WHilE tHis apProaCH taKes compLetely intO acCOunt ThE lOcal intEractions goveRnIng the dynaMiCs oF $\textbF{a}$, It requirEs closures for both the preSSure HesSiaN and vIscoUs term. As fAr aS we knoW, prOVided cLosureS miss At LeaST PartiALLy The NoNlocal natuRE Of pRessuRe [@cheA2]. an alterNative approach woulD Be dEvoted to consiDer The sPATiAl dIStRIbuTiON of THE vectorial velocIty field. VeRy FEw TheoreticaL WorKs Have focUsed on tHis diFFicult, aLthough imPortant, asPeCt. NoNETheLess, it has bEen shown That takinG Into aCCoUnt nuMerIcally ThE shOrt tiMe adveCTioN of flUid parTiCles foR all

, Phys. Rev. [**D67**]{},094002 (20 03),Phy s.Re v. [ **D6 9**]{}, 074026 (200 4) and Phys. Rev. [**D 74**] {} 1140 0 7(2006 ). X. A r tr u , he p- ph /02 07 3 09 (200 2). P. Abb on [*et al .*] {} \[COMPASS c o ll aboration\ ] h ep-ex/070304 9,accept ed by NIM [ **A **]{} . --- author : - 'Laur en t Chevi l lard, R a o ul Rob ert and Vincent V a rg a s' title: A St ochast ic Re p r ese nta tion of th eLocal Structu r eo f Tur b ulence --- I ntroduction === ====== == = F ully d evelo pe d tu rbulent flo ws a re omnipr esenti n Natur e ([e.g. ]{} me teo rol ogy) an deng in e eri n g([e . g.] {} combu st io n). D espi t e a n ap par entcompl exity, it tur nsoutt hat thes e flo ws e xh ibituniver sal s ta tistical proper ties such asthe K olm og orov$ k^{-5/ 3}$ -la w obser ved ont hepo w e r s pectrum [@K41] and t h e i ntermitt ent na t ur eo f longit ud ina l an d trans vers e v elocityfluctu a ti on s [@Fri 95 ]. Mod er n [ dev elopm e nts] {} ofexperime ntala nd numerical f a cilities [@Ts i no 0 1 ;@ Wal0 9]have furthe rmor e und erli n ed th e pecu liaran d u n iversal geometry of t urbule nce. From a mathe matical vi e w p oint, ou r kn o wl e dge of Navier- Stoke s and Eule r equatio ns is still p oor and t h i s makesthe si tua tio n fa r from cleari f welo ok fora r elevant mo del of tu rb ulent flo ws. [Ind ee d, o ne ca n pro v e the ex is ten ce of diss i pative solu tion sof Eul er equa t io n s and t he ass oci at ed in erti a l d issipat ion proce ssh as b ee nstudied [@DucRob00]. A s a conseq ue nce the i n t roductio n of viscosity is not t h e onlyway we h aveto buildrea listic mo d el ofturbul ent f lo ws. ] { } Som e pr ogr es ses have b e e n a chiev ed rec ently i n the understandin g of the universa lit y of t he sm a ll sca le s of t urbulence [by]{ } studying t h eLagrangian dyn am ics ofthe vel ocity gradien t tensor$A_{ij} = \ part i a l_j u_i$, wher e $\text bf{u}$ de n otest he velo cit y (see [ @Ch eA2]and re f ere ncestherei n) . Whil e thi sapproach takes completely intoaccoun t the lo cal inter act i ons governin g th e dynamics of $\ textb f{A } $, it req u ir esc losur es f o r both th e p res s u re Hessian an d v isc ous t erm . As fa r as we know, provide d closures miss atl e ast pa r tial ly the nonlocalnat ur e of press ur e [@CheA2]. An alte rn a tiveapproa ch wou ld be d e v ot e d to c onsi der the spat ial d i stribut io no f thevect or ial ve locity fiel d . Very few theore tical w orksh ave focu se d on th i s di fficult, a lthough imp ortant , as pect. Noneth el ess, i t h as been show n that tak ing i nto acc ou nt n ume ricall y th e short tim eadv ection of f lu i dpa r tic lesfor a ll

, Phys._Rev. [**D67**]{},_094002 (2003), Phys. Rev._[**D69**]{}, 074026_(2004)_and Phys._Rev._[**D74**]{} 114007 (2006)._X. Artru, hep-ph/0207309_(2002). P. Abbon [*et al.*]{}_\[COMPASS collaboration\] hep-ex/0703049,_accepted_by NIM [**A**]{}. --- author: - 'Laurent Chevillard, Raoul Robert and Vincent Vargas' title: A Stochastic Representation_of_the Local_Structure_of_Turbulence --- Introduction ============ Fully developed turbulent flows are_omnipresent in Nature ([e.g.]{} meteorology)_and engineering_([e.g.]{} combustion). Despite an apparent complexity, it turns_out_that these flows_exhibit universal statistical properties such as the Kolmogorov $k^{-5/3}$-law_observed on the power spectrum [@K41]_and the intermittent_nature_of_longitudinal and transverse velocity_fluctuations [@Fri95]. Modern [developments]{} of experimental_and numerical facilities [@Tsino01; @Wal09] have_furthermore underlined the peculiar and universal geometry_of turbulence. From a mathematical viewpoint, our_knowledge of Navier-Stokes and Euler_equations is_still poor and this makes_the situation far_from clear_if we look_for a relevant model of turbulent_flows. [Indeed, one_can prove the existence of dissipative_solutions_of Euler equations_and_the_associated inertial_dissipation process has_been_studied [@DucRob00]._As_a consequence the introduction of viscosity_is_not the only way we have to_build realistic model of_turbulent_flows.]{} Some progresses have_been achieved recently in the_understanding of the universality of the_small scales_of turbulence_[by]{} studying the Lagrangian dynamics of the velocity gradient tensor $A_{ij}_= \partial_ju_i$, where $\textbf{u}$ denotes the_velocity (see [@CheA2] and_references therein)._While_this approach takes_completely_into account_the local interactions governing the dynamics of_$\textbf{A}$, it_requires closures for both the pressure_Hessian and viscous term._As_far as we know, provided closures_miss at least partially the nonlocal_nature of pressure [@CheA2]. An_alternative_approach_would be devoted to consider_the spatial distribution of the vectorial_velocity field. Very_few theoretical works have focused on this_difficult,_although important, aspect. Nonetheless, it has_been_shown that taking into account numerically_the_short_time advection of fluid particles_for all

Sinha and P.Roy, Jour.Phys. [**A40**]{} (2007) 10599\ A.Sinha and P.Roy, Jour.Phys. [**A 42**]{} (2009) 052002\ B.Midya, P.P.Dube and R.Roychoudhury, Jour.Phys. [**A44**]{} (2011) 062001 (FTC) 20. C.M.Bender, Rep.Prog.Phys. [**70**]{} (2007) 947\ A.Mostafazadeh, Phys.Scrip. [**82**]{}(2010) 038110 21. F.G.Scholtz, H.B.Geyer and J.Hahne, Ann.Phys. [**213**]{} (1992) 74\ A.Mostafazadeh and J.Batal, Jour.Phys. [**A37**]{} (2004) 11645\ R.Kretschmer and L.Szymanowski, Phys.Letts. [**A325**]{} (2004) 112 22. B.Bagchi, C.Quesne, R.Roychoudhury, Jour.Phys. [**A38**]{} (2005) L647 23. L.M.Yang, Phys.Rev. [**84**]{} (1951) 788\ H.L.Carrion and R.De Lima Rodrigues, Mod.Phys.Letts. [**A 25**]{} (2010) 2507 24. V.de Alfaro, S.Fubini and G.Furlan, Nuovo Cimento [**A34**]{} (1976) 569 25. E.Witten, Nucl.Phys.B [**185**]{} (1981) 513\ E.Witten, Nucl.Phys.B [**202**]{} (1982) 253 26. S.Fubini and E.Rabinovici, Nucl.Phys.B [**245**]{} (1984) 17 27. R.Haag, V.T.Lopuszanski, M.Sohnius, Nucl.Phys.B [88]{} (1975) 257\ W.Nahm, V.Rittenberg, M.Scheunet, Phys.Letts. [**B61**]{} (1976) 383 28. F.Cannata, G.J

Sinha and P.Roy, Jour. Phys. [ * * A40 * * ] { } (2007) 10599\ A.Sinha and P.Roy, Jour. Phys. [ * * A 42 * * ] { } (2009) 052002\ B.Midya, P.P.Dube and R.Roychoudhury, Jour. Phys. [ * * A44 * * ] { } (2011) 062001 (FTC) 20. C.M.Bender, Rep. Prog. Phys. [ * * 70 * * ] { } (2007) 947\ A.Mostafazadeh, Phys. Scrip. [ * * 82**]{}(2010) 038110 21. F.G.Scholtz, H.B.Geyer and J.Hahne, Ann. Phys. [ * * 213 * * ] { } (1992) 74\ A.Mostafazadeh and J.Batal, Jour. Phys. [ * * A37 * * ] { } (2004) 11645\ R.Kretschmer and L.Szymanowski, Phys. Letts. [ * * A325 * * ] { } (2004) 112 22. B.Bagchi, C.Quesne, R.Roychoudhury, Jour. Phys. [ * * A38 * * ] { } (2005) L647 23. L.M.Yang, Phys. Rev. [ * * 84 * * ] { } (1951) 788\ H.L.Carrion and R.De Lima Rodrigues, Mod. Phys. Letts. [ * * A 25 * * ] { } (2010) 2507 24. V.de Alfaro, S.Fubini and G.Furlan, Nuovo Cimento [ * * A34 * * ] { } (1976) 569 25. E.Witten, Nucl. Phys. B [ * * 185 * * ] { } (1981) 513\ E.Witten, Nucl. Phys. B [ * * 202 * * ] { } (1982) 253 26. S.Fubini and E.Rabinovici, Nucl. Phys. B [ * * 245 * * ] { } (1984) 17 27. R.Haag, V.T.Lopuszanski, M.Sohnius, Nucl. Phys. B [ 88 ] { } (1975) 257\ W.Nahm, V.Rittenberg, M.Scheunet, Phys. Letts. [ * * B61 * * ] { } (1976) 383 28. F.Cannata, G.J

Sinja and P.Roy, Jour.Phys. [**A40**]{} (2007) 10599\ A.Sinha and K.Riy, Jouc.Phys. [**A 42**]{} (2009) 052002\ B.Oidya, P.P.Dube and R.Roychoudhucy, Jiur.Phts. [**A44**]{} (2011) 062001 (FTC) 20. C.M.Bender, Reo.Prog.Phys. [**70**]{} (2007) 947\ A.Nostefazadeh, Phys.Scrml. [**82**]{}(2010) 038110 21. F.G.Sgkoltz, M.B.Geyzr and J.Hahne, Ank.Phys. [**213**]{} (1992) 74\ A.Mostafazadeh avd J.Batal, Jour.Phys. [**A37**]{} (2004) 11645\ R.Kretschmer and L.Saylanowski, Phys.Lgtts. [**S325**]{} (2004) 112 22. B.Gagchi, C.Quesne, R.Roychoudhury, Jour.Pgys. [**A38**]{} (2005) N647 23. L.M.Yang, Phys.Tev. [**84**]{} (1951) 788\ H.L.Carrion and R.Df Lila Rodrigues, Mod.Phjs.Letts. [**A 25**]{} (2010) 2507 24. V.dq Alfaro, S.Fubkni and G.Flxlan, Nuovo Dimento [**A34**]{} (1976) 569 25. E.Witten, Nucl.Phys.B [**185**]{} (1981) 513\ Z.Witten, Nuco.Pyys.H [**202**]{} (1982) 253 26. S.Fubini and V.Rabinovici, Nmbl.Phys.B [**245**]{} (1984) 17 27. R.Hasg, V.T.Lopuszansli, K.Soynius, Nucl.Phys.B [88]{} (1975) 257\ W.Nahm, V.Rittenberg, M.Scheuned, 'hys.Letts. [**B61**]{} (1976) 383 28. F.Cannatq, T.J

Sinha and P.Roy, Jour.Phys. [**A40**]{} (2007) 10599\ P.Roy, [**A 42**]{} 052002\ B.Midya, P.P.Dube 062001 20. C.M.Bender, Rep.Prog.Phys. (2007) 947\ A.Mostafazadeh, [**82**]{}(2010) 038110 21. F.G.Scholtz, H.B.Geyer and Ann.Phys. [**213**]{} (1992) 74\ A.Mostafazadeh and J.Batal, Jour.Phys. [**A37**]{} (2004) 11645\ R.Kretschmer and Phys.Letts. [**A325**]{} (2004) 112 22. B.Bagchi, C.Quesne, R.Roychoudhury, Jour.Phys. [**A38**]{} (2005) L647 23. Phys.Rev. (1951) H.L.Carrion R.De Lima Rodrigues, Mod.Phys.Letts. [**A 25**]{} (2010) 2507 24. V.de Alfaro, S.Fubini and G.Furlan, Nuovo Cimento (1976) 569 25. E.Witten, Nucl.Phys.B [**185**]{} (1981) 513\ Nucl.Phys.B [**202**]{} (1982) 253 S.Fubini and E.Rabinovici, Nucl.Phys.B [**245**]{} 17 R.Haag, V.T.Lopuszanski, Nucl.Phys.B (1975) W.Nahm, V.Rittenberg, M.Scheunet, [**B61**]{} (1976) 383 28. F.Cannata, G.J

Sinha and P.Roy, Jour.Phys. [**A40**]{} (2007) 10599\ A.SinHa and P.Roy, JOur.PhYs. [**A 42**]{} (2009) 052002\ b.MiDyA, P.P.DUbe aNd R.RoychoudhurY, jour.phys. [**A44**]{} (2011) 062001 (FTC) 20. C.M.Bender, Rep.ProG.Phys. [**70**]{} (2007) 947\ a.MOStafAZaDeh, PhYs.Scrip. [**82**]{}(2010) 038110 21. f.g.SCHOltZ, H.b.GEyeR aND J.hahne, ann.phys. [**213**]{} (1992) 74\ A.MoStafazadeh And j.BAtal, Jour.Phys. [**a37**]{} (2004) 11645\ r.KRetschmer aNd L.szymanowski, PHys.letts. [**A325**]{} (2004) 112 22. b.BAgcHI, C.QueSne, r.RoycHoudhuRY, Jour.PHys. [**A38**]{} (2005) L647 23. L.M.YaNg, pHys.Rev. [**84**]{} (1951) 788\ h.l.CarrioN ANd r.De LIma Rodrigues, Mod.PhYS.LETts. [**A 25**]{} (2010) 2507 24. V.de Alfaro, S.fubini AnD g.FURLan, nuoVo Cimento [**A34**]{} (1976) 569 25. e.WItten, nUcl.Phys.b [**185**]{} (1981) 513\ e.WITTEn, NUCl.Phys.B [**202**]{} (1982) 253 26. S.FubinI and E.RabinoVIci, nucl.PhYs.b [**245**]{} (1984) 17 27. R.HAAg, V.T.LoPuszaNsKI, M.SOhnius, Nucl.PHys.B [88]{} (1975) 257\ w.Nahm, V.RitTenberG, m.ScheunET, Phys.LeTts. [**B61**]{} (1976) 383 28. F.CAnnAta, g.J

Sinha and P.Roy, Jour.Phys . [**A40** ]{} ( 200 7)10 599\ A.Sinha and P . Roy, Jour.Phys. [**A 42**] {} (2 00 9 ) 05 2 00 2\ B.Mid y a, P .P. Du be an dR .R oycho udh ury, Jo ur.Phys. [ **A 44 **]{} (2011) 06 2001 (FTC) 2 0. C.M.Bende r,Rep.Pr og .Ph y s. [* *70 **]{} (2007 ) 947\ A.Mos ta f azadeh , Phys.S c r ip . [* *82**]{}(2010) 03 8 11 0 21. F.G.Scho ltz, H .B . Ge y e r a ndJ.Hahne, A nn .Phys . [**213 * *] { } (19 9 2) 74\ A. Mostafazade h an d J.Ba ta l,J our.Ph ys. [ ** A 37* *]{} (2004) 116 45\ R .Krets c hmer an d L.Szym anowsk i,Phy s.Le t ts .[** A3 2 5** ] {} (2 0 04) 112 22 .B. Bagch i, C . Q u e sne, R. Royc houdh ury, Jour.Phy s.[**A 3 8** ]{} ( 2005) L64 7 23.L.M.Ya ng, P hy s.Rev. [**84**] {} ( 1951) 788 \ H. L. Carri o n andR.D e L ima Rod rigues, Mod .P h y s .L etts. [**A 25**]{} ( 2 0 10 ) 2507 24. V. d eAl f aro, S.F ub ini and G .Furl an,N uo vo Cimen to [** A 34 ** ]{} (19 76 ) 569 2 5.E.W itten , Nuc l.Phys .B [**18 5**]{ } (1981) 513\ E.Witten, N u cl . P hy s .B [ **2 02**]{} (19 82)2 53 26.S .F ubi n i and E.Ra bi n ov i ci, Nucl.Phys.B [** 24 5**]{} (198 4) 17 27. R. Haag, V.T. L o p uszanski , M. S oh n ius, Nucl.Phys .B [8 8]{} (1975 ) 257\ W.N ahm, V.R ittenberg , M.Scheun et, Ph ys. Let t s .[**B61**]{} ( 1 9 76)38 3 28.F.C annata, G. J

Sinha and_P.Roy, Jour.Phys._[**A40**]{} (2007) 10599\ _ _A.Sinha_and P.Roy,_Jour.Phys._[**A 42**]{} (2009)_052002\ _ B.Midya, P.P.Dube and_R.Roychoudhury, Jour.Phys. [**A44**]{}_(2011)_062001 (FTC) 20. C.M.Bender, Rep.Prog.Phys. [**70**]{} (2007) 947\ A.Mostafazadeh, Phys.Scrip. [**82**]{}(2010) 038110 21._F.G.Scholtz,_H.B.Geyer and_J.Hahne,_Ann.Phys._[**213**]{} (1992) 74\ _ A.Mostafazadeh and J.Batal, Jour.Phys._[**A37**]{} (2004)_11645\ R.Kretschmer and L.Szymanowski, Phys.Letts._[**A325**]{}_(2004) 112 22. B.Bagchi,_C.Quesne, R.Roychoudhury, Jour.Phys. [**A38**]{} (2005) L647 23. L.M.Yang, Phys.Rev. [**84**]{}_(1951) 788\ H.L.Carrion_and R.De Lima_Rodrigues,_Mod.Phys.Letts._[**A 25**]{} (2010) 2507 24._V.de Alfaro, S.Fubini and G.Furlan, Nuovo_Cimento [**A34**]{} (1976) 569 25. E.Witten, Nucl.Phys.B_[**185**]{} (1981) 513\ E.Witten,_Nucl.Phys.B [**202**]{} (1982) 253 26. S.Fubini and_E.Rabinovici, Nucl.Phys.B [**245**]{} (1984) 17 27._R.Haag, V.T.Lopuszanski,_M.Sohnius, Nucl.Phys.B [88]{} (1975) 257\ _ _W.Nahm, V.Rittenberg,_M.Scheunet, Phys.Letts. [**B61**]{}_(1976) 383 28. F.Cannata, G.J

supporting observational evidence has so far failed to materialize. Interestingly, some recent observations hint at dark matter beyond this narrow GeV-TeV window, indicating that the structure of dark matter and BSM physics might be very different from what was envisioned. One such observation is the IceCube neutrino observatory’s detection of 37 neutrino events in the energy range from 30 GeV to 2 PeV, disfavoring a purely atmospheric explanation at $5.7 \sigma$ [@Aartsen:2014gkd]. In particular, IceCube has reported 3 neutrino events above 1 PeV. Strongly incompatible with traditional astrophysical processes, these have been shown to be compatible with the decay of heavy, long lived particles with lifetimes of about 10$^{27}$s that constitute some or all of dark matter [@Fong:2014bsa; @Ema:2013nda; @Ema:2014ufa; @Feldstein:2013kka; @Esmaili:2013gha; @Esmaili:2014rma; @Bai:2013nga; @Bhattacharya:2014vwa; @Higaki:2014dwa; @Bhattacharya:2014yha; @Rott:2014kfa]. A dark matter interpretation of these events, however, poses several theoretical inconveniences: Why should we expect new physics at the PeV scale? Why should a PeV scale particle be so long-lived? What sets its relic density? Why should it show up in neutrinos? Given the dearth of BSM candidates at the PeV scale, a theoretically motivated framework that also fits into the broader particle physics picture would be extremely appealing. Another observation, made independently by two groups, is the discovery of an unidentified X-ray line at $E_\gamma \approx 3.5$ keV in the stacked X-ray spectra of 73 galaxy clusters measured by XMM-Newton [@Bulbul:2014sua] and in the X-ray spectra of the Andromeda galaxy and the Perseus galaxy cluster [@Boyarsky:2014jta]. A dark matter interpretation of these observations is plagued by many unresolved questions, such as possible contamination from potassium lines [@Jeltema:2014qfa; @Boyarsky:2014paa; @Bulbul:2014ala] as well as inconsistency with stacked observations of dwarf spheroidals [@Malyshev:2014xqa] and galaxy spectra [@

supporting observational evidence has so far failed to happen. Interestingly, some late notice hint at dark topic beyond this narrow GeV - TeV window, indicating that the social organization of colored matter and BSM purgative might be very unlike from what was envisioned. One such observation is the IceCube neutrino lookout ’s detection of 37 neutrino events in the department of energy range from 30 GeV to 2 PeV, disfavoring a purely atmospheric explanation at $ 5.7 \sigma$ [ @Aartsen:2014gkd ]. In particular, IceCube has reported 3 neutrino events above 1 PeV. Strongly antagonistic with traditional astrophysical processes, these have been shown to be compatible with the decay of heavy, long exist particles with lifetimes of about 10$^{27}$s that constitute some or all of dark topic [ @Fong:2014bsa; @Ema:2013nda; @Ema:2014ufa; @Feldstein:2013kka; @Esmaili:2013gha; @Esmaili:2014rma; @Bai:2013nga; @Bhattacharya:2014vwa; @Higaki:2014dwa; @Bhattacharya:2014yha; @Rott:2014kfa ]. A dark topic interpretation of these events, however, poses several theoretical troublesomeness: Why should we expect new physics at the PeV scale? Why should a PeV scale atom be thus long - lived? What sets its relic density? Why should it show up in neutrinos? Given the dearth of BSM candidates at the PeV scale, a theoretically motivated model that besides fits into the broader particle purgative movie would be extremely appealing. Another observation, made independently by two groups, is the discovery of an unidentified X - re line at $ E_\gamma \approx 3.5 $ keV in the stacked X - ray spectra of 73 galaxy clusters measured by XMM - Newton [ @Bulbul:2014sua ] and in the adam - ray spectra of the Andromeda galaxy and the Perseus galaxy cluster [ @Boyarsky:2014jta ]. A dark topic interpretation of these observation is blight by many unresolved questions, such as possible contamination from potassium line [ @Jeltema:2014qfa; @Boyarsky:2014paa; @Bulbul:2014ala ] as well as inconsistency with stacked observation of dwarf spheroidals [ @Malyshev:2014xqa ] and galaxy spectra [ @

suoporting observational eyidence has so fce failxd to mzterialixe. Interestingly, some recent ovservqtions hint at dark magter beyojd this barriq GeV-TeV wmhdow, indicatihn thac vhe structure on dark mattar and BSM phyvizs might be very different from what ras envosloned. One such jbsegvwtioh is the IceCube neutrino observatkry’s deuection of 37 neutrimo events in the energy rajge vrom 30 GeV to 2 PeV, fisfavoring a pteely atmosphdric explanation at $5.7 \sjgma$ [@Aartsen:2014gkd]. In particular, IzeCubz has reporjze 3 jgutrino evenvs abone 1 PeV. Stronnky incmmpatibke with traditlonal asrrophysical processes, these have been srown to ba eompatible with the dwcqy of heaey, lubg uivtd 'arficles wivh lifetimea of about 10$^{27}$w that constitute spmq or all of dadk mateew [@Fong:2014bsa; @Ema:2013nda; @Ema:2014ufa; @Feldstein:2013kka; @Eskaimi:2013gha; @Esmaili:2014rma; @Bai:2013nga; @Bhattacharya:2014vwa; @Higani:2014dwa; @Bhaetacharya:2014yha; @Rott:2014kfa]. A dark matter interpretation of tiere tvckts, fiwfver, poses several theoretical inconveniences: Shu xhould we expegt new physics at yhf LgV scale? Why sfould c PsV scale particle he so ljng-liced? What fets its relic density? Why shouod it show ui in neutrinos? Given tke dearth of BSM candodates at the PeV scale, a tgeoreticallj motivatsa framework that alxo fits inuu the broader pareicle phywics picturd woold be qxtremely wppealing. Another observahion, lage indepenfently by two groups, is the disrpvery of an onigennified X-rcy linc at $E_\gamma \appwox 3.5$ keV in thg stacked X-ray spectra or 73 galaey clusters ieasured by XKL-Newton [@Bulbnl:2014sua] and in rhe Z-ray spdztra of the Ancromeda gcjaxy and the Perseus gelaxi dluster [@Boyarskv:2014ltq]. A dark matter inger[rvtavion jx these obsesvatkonr is puagued by mcnh untesolved questions, sgch zs possible contamonwtion frim potasfium lines [@Jektema:2014qfa; @Boyarsky:2014pwa; @Bunbuu:2014ala] ss rell as inconsistency with stadked obsegvabions of dwars spmeroydals [@Malyxhev:2014xqa] and galaxy spectra [@

supporting observational evidence has so far failed Interestingly, recent observations at dark matter indicating the structure of matter and BSM might be very different from what envisioned. One such observation is the IceCube neutrino observatory’s detection of 37 neutrino in the energy range from 30 GeV to 2 PeV, disfavoring a purely explanation $5.7 [@Aartsen:2014gkd]. particular, IceCube has reported 3 neutrino events above 1 PeV. Strongly incompatible with traditional astrophysical processes, have been shown to be compatible with the of heavy, long lived with lifetimes of about 10$^{27}$s constitute or all dark [@Fong:2014bsa; @Ema:2014ufa; @Feldstein:2013kka; @Esmaili:2013gha; @Bai:2013nga; @Bhattacharya:2014vwa; @Higaki:2014dwa; @Bhattacharya:2014yha; @Rott:2014kfa]. A dark matter interpretation of these events, however, poses several theoretical inconveniences: should we physics at PeV Why a PeV scale so long-lived? What sets its relic it show up in neutrinos? Given the dearth BSM candidates the PeV scale, a theoretically motivated that also fits into the broader particle physics would be extremely appealing. Another observation, made independently by two groups, is the discovery of X-ray line at $E_\gamma 3.5$ keV in stacked spectra 73 clusters measured XMM-Newton [@Bulbul:2014sua] and in the X-ray spectra of the Andromeda galaxy the Perseus galaxy cluster [@Boyarsky:2014jta]. A dark matter interpretation of is by many unresolved such as possible contamination potassium [@Jeltema:2014qfa; @Boyarsky:2014paa; @Bulbul:2014ala] as inconsistency stacked spheroidals and spectra [@

supporting observational evIdence has sO far fAilEd tO mAterIaliZe. InterestinglY, Some Recent observations hint At darK mATter BEyOnd thIs narroW gev-tEV wInDoW, inDiCAtIng thAt tHe strucTure of dark MatTeR and BSM physiCS mIght be very DifFerent from whAt wAs enviSiOneD. one suCh oBservAtion iS The Icecube neutrInO ObservATory’s deTECtIon oF 37 neutrino events in THe ENergy range from 30 geV to 2 PEV, DIsFAVorIng A purely atmOsPheriC ExplanaTIoN AT $5.7 \SigMA$ [@Aartsen:2014gkd]. In Particular, ICECuBe has rEpOrtED 3 neutrIno evEnTS abOve 1 PeV. StronGly iNcompatibLe with TRaditioNAl astroPhysicAl pRocEsseS, ThEsE haVe BEen SHoWn tO Be cOmpatiblE wItH the dEcay OF HEAvy, lOng LiveD partIcles with lifeTimEs of ABouT 10$^{27}$s thaT consTituTe Some oR all of Dark mAtTer [@Fong:2014bsa; @Ema:2013ndA; @Ema:2014Ufa; @FeldstEin:2013KkA; @EsMaIli:2013ghA; @esmailI:2014rmA; @BaI:2013nga; @BhaTtacharYA:2014vwA; @HIGAKi:2014Dwa; @Bhattacharya:2014yha; @roTT:2014KfA]. A dark maTter inTErPrETation of ThEse EvenTS, HowevEr, poSEs Several tHeoretICaL iNconvenIeNces: WhY sHouLd wE expeCT new PhysicS at the Pev scalE? why should a PeV sCAle particle be SO lONG-lIVed? WHat Sets its reliC denSIty? WHy shOUlD it SHow up In neuTrINoS? given the dearth of BSM CaNdidatEs at tHe PeV scale, a thEoreticallY MOTivated fRameWOrK That also fits inTo the Broader parTIcle physIcs piCture wouLd be extreMELy appealIng. anoTheR obSERvAtion, made indePENdenTlY by two gRouPs, is the DisCovEry Of aN uNidentifiEd X-ray liNe At $e_\gAmMa \aPprox 3.5$ KEV in the sTaCkeD X-Ray SpectRA of 73 galAxy clUsteRs MeASurEd by XMM-nEwTON [@BulBuL:2014sUa] anD in ThE X-ray SpecTRa oF the AndRomeda galAxy ANd thE PErSeus galAxy cluster [@BoyArSky:2014jta]. A darK mAttEr inteRPRetation Of these observations is plAGued by mAny UnresOlveD questionS, suCh as poSsiBLe contAminatIon frOm PotASSium lINEs [@jelTeMa:2014qfa; @BoyarSKY:2014paA; @BulbUl:2014Ala] aS well as Inconsistency with sTAckEd observationS of DwarF SPhEroIDaLS [@MaLySHev:2014XQA] and galaxy spectRa [@

supporting observationalevidence h as so fa r f ai ledto m aterialize. In t eres tingly, some recent ob serva ti o ns h i nt at d ark mat t er b eyo nd t his n a rr ow Ge V-T eV wind ow, indica tin gthat the str u ct ure of dar k m atter and BS M p hysics m igh t be v ery diff erentf rom wh at was en vi s ioned. One su c h o bser vation is the Ice C ub e neutrino obse rvator y’ s d e t ect ion of 37 neu tr ino e v ents in th e e ner g y range from30 GeV to 2 PeV , disf av ori n g a pu relyat m osp heric expla nati on at $5. 7 \sig m a$ [@Aa r tsen:20 14gkd] . I n p arti c ul ar , I ce C ube ha s r e por ted 3 ne ut ri no ev ents a b o ve 1 Pe V. S trong ly incompatib lewith tra ditio nal a stro ph ysica l proc esses ,these have been sho wn to becom pa tib le with the de cay of heavy, long l i ved p a r t ic les with lifetimes o f ab out 10$^ {27}$s th at constitu te so me o r all o f da r kmatter [ @Fong: 2 01 4b sa; @Em a: 2013nd a; @E ma: 2014u f a; @ Feldst ein:2013 kka;@ Esmaili:2013gh a ; @Esmaili:20 1 4r m a ;@ Bai: 201 3nga; @Bhat tach a rya: 2014 v wa ; @ H igaki :2014 dw a ;@ Bhattacharya:2014yh a; @Rott :2014 kfa]. A darkmatter int e r p retation oft he s e events, howe ver,poses seve r al theor etica l inconv eniences: W hy shoul d w e e xpe ctn e wphysics at th e PeVsc ale? Wh y s hould a Pe V s cal e p ar ticle beso long- li ve d? W hat sets its reli cden si ty? Whys houldit sh ow u pin neu trinos? Gi v e n th ede arth of B SM ca ndid a tes at the PeV scal e,a the or et icallymotivated fra me work thatal sofits i n t o the br oader particle physicsp icturewou ld be ext remely ap pea ling. An o ther o bserva tion, m ade i ndepe n d en tly b y two grou p s , i s the d isco very of an unidentified X - ray line at $E_\ gam ma \ a p pr ox3 .5 $ ke Vi n t h e stacked X-rayspectra of 7 3 g alaxy clus t ers m easured by XMM -Newt o n [@Bul bul:2014s ua] and i ntheX - ray spectra o f the An dromeda g a laxya nd thePer seus g al axy clus ter [@ B oya rsky: 2014jt a] . A da rk ma tt er inter pretation of these obse rvatio ns is pl agued byman y un resolvedques tions, suc h a s p ossib lec ontam inat i on fr o m pot assi u m lines [ @ Je lte m a :2 014qfa; @Bo y a r sky :2014 paa ; @Bulb ul:2 014ala] as well a s inconsistency wit h sta cke d obs er vations of dwa rfsp h e roidals[@ Malyshev:20 14xqa] a nd galax y spec tra [@

supporting_observational evidence_has so far failed_to materialize._Interestingly,_some recent_observations_hint at dark_matter beyond this_narrow GeV-TeV window, indicating_that the structure_of_dark matter and BSM physics might be very different from what was envisioned. One such_observation_is the_IceCube_neutrino_observatory’s detection of 37 neutrino_events in the energy range_from 30_GeV to 2 PeV, disfavoring a purely atmospheric_explanation_at $5.7 \sigma$_[@Aartsen:2014gkd]. In particular, IceCube has reported 3 neutrino events_above 1 PeV. Strongly incompatible with_traditional astrophysical processes,_these_have_been shown to be_compatible with the decay of heavy,_long lived particles with lifetimes of_about 10$^{27}$s that constitute some or all_of dark matter [@Fong:2014bsa; @Ema:2013nda; @Ema:2014ufa;_@Feldstein:2013kka; @Esmaili:2013gha; @Esmaili:2014rma; @Bai:2013nga; @Bhattacharya:2014vwa;_@Higaki:2014dwa; @Bhattacharya:2014yha;_@Rott:2014kfa]. A dark matter interpretation_of these events,_however, poses_several theoretical inconveniences:_Why should we expect new physics_at the PeV_scale? Why should a PeV scale_particle_be so long-lived?_What_sets_its relic_density? Why should_it_show up_in_neutrinos? Given the dearth of BSM_candidates_at the PeV scale, a theoretically motivated_framework that also fits_into_the broader particle physics_picture would be extremely appealing. Another_observation, made independently by two groups,_is the_discovery of_an unidentified X-ray line at $E_\gamma \approx 3.5$ keV in the_stacked X-ray spectra of 73 galaxy_clusters measured by XMM-Newton_[@Bulbul:2014sua] and_in_the X-ray spectra_of_the Andromeda_galaxy and the Perseus galaxy cluster [@Boyarsky:2014jta]._A dark_matter interpretation of these observations is_plagued by many unresolved_questions,_such as possible contamination from potassium_lines [@Jeltema:2014qfa; @Boyarsky:2014paa; @Bulbul:2014ala] as well_as inconsistency with stacked observations_of_dwarf_spheroidals [@Malyshev:2014xqa] and galaxy spectra_[@

topograph and (b) derivative image of a graphene flake on Si(111). The atomic arrangement of Si(111)$-$7$\times$7 can be clearly seen underneath graphene at the inset. (c) An empty states STM image and its derivative image help establishing that the 7$\times$7 reconstruction of Si(111) remains unchanged after graphene was adsorbed.](Fig2sm.pdf){width="40.00000%"} To this end, we interface the stable Si(111)$-$7$\times$7 surface with graphene using the dry-contact transfer technique, [@Albrecht1] as illustrated in Fig. 1(a). Our transfer process produces a clean interface at the graphene-Si(111) heterostructure, allowing examination with atomically resolved, ultrahigh vacuum (UHV) scanning tunneling microscopy (STM) and spectroscopy (STS). We frame these experimental results and explore the interfacial chemistry by [*ab-initio*]{} density-functional theory (DFT) calculations using a large supercell in the local density approximation (LDA). These calculations reproduce our experimental findings and reveal the absence of covalent C$-$Si bonds at the graphene-Si(111) interface. Our article complements our previous STM studies of the atomic-scale interaction of carbon-based materials and technologically-relevant silicon substrates.[@Albrecht1; @Orellana; @E1; @E2; @E3; @E4; @E5; @E6; @E8] In the approach of Fig. 1(a), a fiberglass applicator collects graphene flakes and gently presses them onto a hydrogen-free Si(111) substrate, allowing for subsequent, [*in situ*]{}, room-temperature STM interrogation. We note that the orientation of graphene flakes from the applicator is random, and, hence, we expect the registry between the graphene and Si(111) lattices to be random also. A filled-states STM image of two graphene flakes on Si(111) is displayed on Fig. 1(b); the graphene flakes are seen on a lighter color on the image. Despite the unknown relative orientation between graphene and Si(111), the diamond-like structure characteristic of Si(111)$-$$7\times 7$ can be seen throughout the image. We superimpose two diamond-like features on the graphene and onto the bare Si(111) surface to help make the area spanned by Si(111) reconstructed supercells more evident.

topograph and (b) derivative image of a graphene flake on Si(111). The atomic arrangement of Si(111)$-$7$\times$7 can be intelligibly see underneath graphene at the inset. (c) An empty states STM image and its derivative picture help establishing that the 7$\times$7 reconstruction of Si(111) persist unaltered after graphene was adsorbed.](Fig2sm.pdf){width="40.00000% " } To this end, we interface the stable Si(111)$-$7$\times$7 surface with graphene using the dry - contact transportation technique, [ @Albrecht1 ] as illustrated in Fig.   1(a). Our transfer process grow a clean interface at the graphene - Si(111) heterostructure, allowing examination with atomically resolved, ultrahigh vacuum (UHV) scanning tunneling microscopy (STM) and spectroscopy (STS). We ensnare these experimental results and explore the interfacial chemistry by [ * ab - initio * ] { } concentration - functional theory (DFT) calculations using a large supercell in the local concentration approximation (LDA). These calculation reproduce our experimental findings and reveal the absence of covalent C$ -$Si bond at the graphene - Si(111) interface. Our article complements our previous STM studies of the atomic - scale interaction of carbon paper - establish materials and technologically - relevant silicon substrates.[@Albrecht1; @Orellana; @E1; @E2; @E3; @E4; @E5; @E6; @E8 ] In the approach of Fig.   1(a), a fiberglass applicator collects graphene flakes and gently presses them onto a hydrogen - free Si(111) substrate, allowing for subsequent, [ * in situ * ] { }, room - temperature STM interrogation. We notice that the predilection of graphene flakes from the applicator is random, and, therefore, we ask the register between the graphene and Si(111) lattices to be random also. A filled - states STM image of two graphene flakes on Si(111) is displayed on Fig.   1(b); the graphene flakes are see on a lighter color on the image. Despite the unknown relative orientation between graphene and Si(111), the baseball diamond - like structure characteristic of Si(111)$-$$7\times 7 $ can be seen throughout the image. We superimpose two ball field - like feature of speech on the graphene and onto the bare Si(111) surface to help make the area cross by Si(111) reconstructed supercells more evident.

tooograph and (b) derivative image of a grakhwne fleke on Ai(111). The agomic arrangement of Si(111)$-$7$\times$7 cqn be clearly seen underneagh grapheje at thw inwwt. (c) An em'fy statcf STJ imaye and its derivstive imaga help establivhkny that the 7$\times$7 reconstruction of Sy(111) remaims unchanged aftgr grs[hens was adsorbed.](Fig2sm.pdf){width="40.00000%"} To this end, we interface thr stable Si(111)$-$7$\times$7 surface wlth hraphene using the dry-contact tragwfer techniqje, [@Albrechu1] cs illustrajed in Fig. 1(a). Our transfer process prodbces a cleab untftface at the graprene-Si(111) heterostructuse, alloeing examinatipn xith atomically resolved, nltrahigh vacuum (UHV) scanning tbnneling microscopy (SRM) and vpecdroszipy (STA). Xe rrame hheae experimsntal resulrs and explore the onevtfacial chemjstry fy [*ab-initio*]{} density-functional theory (DFT) balchlations using a large wupercell in the locap density approximation (LDA). These calculations reproduce ogr ex'efimtnbwu flndings and reveal the absence of covalent C$-$Sy bpncs at the grapmene-Si(111) interface. Oir atjicle complemevts oux pdevious STM studied of thg atomuc-scale igtersction of carbon-based materuals and tecknooogically-relevant dilicon subrtrajes.[@Albtecht1; @Orellana; @E1; @E2; @E3; @E4; @Z5; @E6; @E8] Ih the approwch of Fif. 1(x), a fiberglass aoplpcatmr collects graphene flakef and genvly pxesses tfem pnto a hydrogen-fgee Sl(111) substrate, allowinh for sgbsequent, [*ln situ*]{}, room-temperature STM intxcrogation. We mode nhat the jrienbation of graphqne flakes frok the a'plicagor is ransom, and, hence, we evpect the reghdtry between the gra[henw ane Si(111) laggices to be ramdom also. A filled-srates STM image of twu graphene flakew ob Si(111) is displayrd un Sih. 1(b); tra graphene fnaker afr seev on a lighber colpr on the image. Desphte fhe unknown relatife orientajion betwqen graphene snd Si(111), the diamond-pike vtrncture chwracteristic of Si(111)$-$$7\times 7$ can bs seen thgounhout the imade. Wc sukerimpose tco diamond-like features on the graphene end onto the bare Si(111) sorface to help make tke area spannxd by Fi(111) reconsdructed supercells mire evident.

topograph and (b) derivative image of a on The atomic of Si(111)$-$7$\times$7 can at inset. (c) An states STM image its derivative image help establishing that 7$\times$7 reconstruction of Si(111) remains unchanged after graphene was adsorbed.](Fig2sm.pdf){width="40.00000%"} To this end, interface the stable Si(111)$-$7$\times$7 surface with graphene using the dry-contact transfer technique, [@Albrecht1] illustrated Fig. Our process produces a clean interface at the graphene-Si(111) heterostructure, allowing examination with atomically resolved, ultrahigh vacuum scanning tunneling microscopy (STM) and spectroscopy (STS). We these experimental results and the interfacial chemistry by [*ab-initio*]{} theory calculations using large in local density approximation These calculations reproduce our experimental findings and reveal the absence of covalent C$-$Si bonds at the graphene-Si(111) Our article previous STM of atomic-scale of carbon-based materials silicon substrates.[@Albrecht1; @Orellana; @E1; @E2; @E3; @E8] In the approach of Fig. 1(a), a applicator collects flakes and gently presses them onto hydrogen-free Si(111) substrate, allowing for subsequent, [*in situ*]{}, STM interrogation. We note that the orientation of graphene flakes from the applicator is random, we expect the registry the graphene and lattices be also. filled-states STM of two graphene flakes on Si(111) is displayed on Fig. 1(b); graphene flakes are seen on a lighter color on the the relative orientation between and Si(111), the diamond-like characteristic Si(111)$-$$7\times 7$ can be the We features the and onto the bare surface to help make the spanned by Si(111) reconstructed

topograph and (b) derivative imAge of a grapHene fLakE on si(111). the aTomiC arrangement of sI(111)$-$7$\timEs$7 can be clearly seen undeRneatH gRApheNE aT the iNset. (c) An EMpTY StaTeS StM iMaGE aNd its DerIvative Image help eStaBlIshing that thE 7$\TiMes$7 reconstRucTion of Si(111) remaIns UnchanGeD afTEr graPheNe was AdsorbED.](Fig2sm.Pdf){width="40.00000%"} TO tHIs end, wE InterfaCE ThE staBle Si(111)$-$7$\times$7 surface WItH Graphene using tHe dry-cOnTAcT TRanSfeR technique, [@alBrechT1] As illusTRaTED In FIG. 1(a). Our transfer Process prodUCes A clean InTerFAce at tHe graPhENe-SI(111) heterostruCturE, allowing ExaminATion witH AtomicaLly resOlvEd, uLtraHIgH vAcuUm (uhV) sCAnNinG TunNeling miCrOsCopy (StM) anD SPECtroScoPy (STs). We frAme these experImeNtal REsuLts anD explOre tHe InterFacial ChemiStRy by [*ab-initio*]{} denSity-FunctionaL thEoRy (DfT) CalcuLAtions UsiNg a Large suPercell IN thE lOCAL dEnsity approximatioN (Lda). thEse calcuLationS RePrODuce our eXpEriMentAL FindiNgs aND rEveal the AbsencE Of CoValent C$-$si Bonds aT tHe gRapHene-SI(111) InteRface. OUr articlE compLEments our previOUs STM studies oF ThE AToMIc-scAle Interaction Of caRBon-bAsed MAtEriALs and TechnOlOGiCAlly-relevant silicon SuBstratEs.[@AlbRecht1; @Orellana; @e1; @E2; @E3; @E4; @E5; @E6; @E8] In tHE APproach oF Fig. 1(A), A fIBerglass applicAtor cOllects graPHene flakEs and Gently prEsses them ONTo a hydroGen-FreE Si(111) SubSTRaTe, allowing for SUBseqUeNt, [*in sitU*]{}, roOm-tempeRatUre sTM IntErRogation. WE note thaT tHe OrIeNtaTion oF Graphene FlAkeS fRom The apPLicatoR is raNdom, AnD, hENce, We expecT ThE REgisTrY bEtweEn tHe GraphEne aND Si(111) LatticeS to be randOm aLSo. A fIlLeD-states sTM image of two GrAphene flakEs On SI(111) is disPLAyed on FiG. 1(b); the graphene flakes are sEEn on a liGhtEr colOr on The image. DEspIte the UnkNOwn relAtive oRientAtIon BETween GRApHenE aNd Si(111), the diaMONd-lIke stRuCturE characTeristic of Si(111)$-$$7\times 7$ cAN be Seen throughouT thE imaGE. we SupERiMPosE tWO diAMOnd-like features On the graphEnE AnD onto the baRE Si(111) SuRface to Help makE the aREa spannEd by Si(111) recOnstructeD sUperCELls More evidenT.

topograph and (b) derivat ive imageof agra phe ne fla ke o n Si(111). The atom ic arrangement of Si(1 11)$- $7 $ \tim e s$ 7 can be cle a rl y see nun der ne a th grap hen e at th e inset. ( c)An empty state s S TM image a ndits derivati veimagehe lpe stabl ish ing t hat th e 7$\ti mes$7 rec on s tructi o n of Si ( 1 11 ) re mains unchanged a f te r graphene wasadsorb ed . ]( F i g2s m.p df){width= "4 0.000 0 0%"} T o t h i s en d , we interfac e the stabl e Si (111)$ -$ 7$\ t imes$7 surf ac e wi th graphene usi ng the dr y-cont a ct tran s fer tec hnique , [ @Al brec h t1 ]asil l ust r at edi n F ig. 1(a) .Ou r tra nsfe r p r oces s p rodu ces a clean interf ace att hegraph ene-S i(11 1) hete rostru cture ,allowing examin atio n with at omi ca lly r esolv e d, ult rah igh vacuum (UHV)s can ni n g tu nneling microscopy ( S T M) and spe ctrosc o py ( S TS). Wefr ame the s e expe rime n ta l result s ande xp lo re thein terfac ia l c hem istry by [ *ab-in itio*]{} dens i ty-functionalt heory (DFT) c a lc u l at i onsusi ng a largesupe r cell int he lo c al de nsity a p pr o ximation (LDA). The se calcu latio ns reproduceour experi m e n tal find ings an d reveal the ab sence of covale n t C$-$Si bond s at the graphene - S i(111) i nte rfa ce. Ou r ar ticle complem e n ts o ur previo usSTM stu die s o f t heat omic-scal e intera ct io nof ca rbon- b ased mat er ial sand tech n ologic ally- rele va nt sil icon su b st r a tes. [@ Al brec ht1 ;@Orel lana ; @E 1; @E2; @E3; @E4 ; @ E 5; @ E6 ;@E8] I n the approac hof Fig. 1( a) , a fiber g l ass appl icator collects graphen e flakes an d gen tlypresses t hem ontoa h y drogen -freeSi(11 1) su b s trate , al low in g for subs e q uen t, [* in sit u*]{},room-temperature S T M i nterrogation. We not e th att he ori en t ati o n of graphene fl akes fromth e a pplicatori s r an dom, an d, henc e, we expectthe regis try betwe en the g rap hene and S i(111) l attices t o be r a nd om al so. A fil le d-s tates STM i m age of t wo gra ph ene fl akeson Si(111) is displayed on Fig. 1 (b); t he gr aph ene flake s a r e s een on aligh ter coloronthe imag e.D espit e th e u nkn o wn re lati v e orienta t io n b e t we en graphene a n d S i(111 ),t he dia mond -like structure c h aracteristic o f Si ( 1 11) $-$ $ 7\ti me s 7$ can be se enth r o ughout t he image. Wesuperimp os e twodiamon d-like featur e s o n the g raph ene and onto th eb are Si( 11 1) surfac e to h elp ma ke the area s panned by Si(111 ) rec o n struc t edsuper ce lls mor e evi dent.

topograph_and (b)_derivative image of a_graphene flake_on_Si(111). The_atomic_arrangement of Si(111)$-$7$\times$7_can be clearly_seen underneath graphene at_the inset. (c)_An_empty states STM image and its derivative image help establishing that the 7$\times$7 reconstruction_of_Si(111) remains_unchanged_after_graphene was adsorbed.](Fig2sm.pdf){width="40.00000%"} To this end,_we interface the stable Si(111)$-$7$\times$7_surface with_graphene using the dry-contact transfer technique, [@Albrecht1] as_illustrated_in Fig. 1(a). Our_transfer process produces a clean interface at the graphene-Si(111)_heterostructure, allowing examination with atomically resolved,_ultrahigh vacuum (UHV)_scanning_tunneling_microscopy (STM) and spectroscopy_(STS). We frame these experimental results_and explore the interfacial chemistry by_[*ab-initio*]{} density-functional theory (DFT) calculations using a_large supercell in the local density_approximation (LDA). These calculations reproduce_our experimental_findings and reveal the absence_of covalent C$-$Si_bonds at_the graphene-Si(111) interface._Our article complements our previous STM_studies of the_atomic-scale interaction of carbon-based materials and_technologically-relevant_silicon substrates.[@Albrecht1; @Orellana;_@E1;_@E2;_@E3; @E4;_@E5; @E6; @E8] In_the_approach of_Fig. 1(a),_a fiberglass applicator collects graphene flakes_and_gently presses them onto a hydrogen-free Si(111)_substrate, allowing for subsequent,_[*in_situ*]{}, room-temperature STM interrogation._We note that the orientation_of graphene flakes from the applicator_is random,_and, hence,_we expect the registry between the graphene and Si(111) lattices to_be random also. A filled-states STM_image of two graphene_flakes on_Si(111)_is displayed on_Fig. 1(b);_the graphene_flakes are seen on a lighter color_on the_image. Despite the unknown relative orientation_between graphene and Si(111),_the_diamond-like structure characteristic of Si(111)$-$$7\times 7$_can be seen throughout the image._We superimpose two diamond-like features_on_the_graphene and onto the bare_Si(111) surface to help make the_area spanned by_Si(111) reconstructed supercells more evident.

} F_{\alpha_j} -F_{\alpha_j} E_{\alpha_i} &= \delta_{i,j} \frac{K_{\alpha_i} -K^{-1}_{\alpha_i}}{q-q^{-1}}, \\ \displaybreak[0] E_{\alpha_i}^2 E_{\alpha_j} +E_{\alpha_j} E_{\alpha_i}^2 &= (q+q^{-1}) E_{\alpha_i} E_{\alpha_j} E_{\alpha_i}, \quad &&\text{if } |i-j|=1, \\ E_{\alpha_i} E_{\alpha_j} &= E_{\alpha_j} E_{\alpha_i}, \,\qquad\qquad &&\text{if } |i-j|>1, \\ F_{\alpha_i}^2 F_{\alpha_j} +F_{\alpha_j} F_{\alpha_i}^2 &= (q+q^{-1}) F_{\alpha_i} F_{\alpha_j} F_{\alpha_i}, \quad\, &&\text{if } |i-j|=1,\\ F_{\alpha_i} F_{\alpha_j} &= F_{\alpha_j} F_{\alpha_i}, \qquad\ \qquad &&\text{if } |i-j|>1.\end{aligned}$$ Let $\U^+$, $\U^0$ and $\U^-$ be the $\Qq$-subalgebra of $\U$ generated by $E_{\alpha_i}$, $K^{\pm 1}_{\alpha_i}$, and $F_{\alpha_i}$ respectively, for $i \in \I$. We introduce the divided power $F^{(a)}_{\alpha_i} = F^a_{\alpha_i}/[a]!$, where $a \ge 0$, $[a] = (q^{a}- q^{-a})/(q-q^{-1})$ and $[a]! = [1][2] \cdots [a]$. Let $\mA =\Z[q,q^{-1}]$. Let $_\mA\U^{+}$ be the $\mA$-subalgebra of $\U^{+}$ generated by $E^{(a)}_{\alpha_i}$ for various $a \ge 0$ and $i \in \I$. Similarly let $_\mA\U^{-}$ be

} F_{\alpha_j } -F_{\alpha_j } E_{\alpha_i } & = \delta_{i, j } \frac{K_{\alpha_i } -K^{-1}_{\alpha_i}}{q - q^{-1 } }, \\ \displaybreak[0 ] E_{\alpha_i}^2 E_{\alpha_j } + E_{\alpha_j } E_{\alpha_i}^2 & = (q+q^{-1 }) E_{\alpha_i } E_{\alpha_j } E_{\alpha_i }, \quad & & \text{if } |i - j|=1, \\ E_{\alpha_i } E_{\alpha_j } & = E_{\alpha_j } E_{\alpha_i }, \,\qquad\qquad & & \text{if } |i - j|>1, \\ F_{\alpha_i}^2 F_{\alpha_j } + F_{\alpha_j } F_{\alpha_i}^2 & = (q+q^{-1 }) F_{\alpha_i } F_{\alpha_j } F_{\alpha_i }, \quad\, & & \text{if } |i - j|=1,\\ F_{\alpha_i } F_{\alpha_j } & = F_{\alpha_j } F_{\alpha_i }, \qquad\ \qquad & & \text{if } |i - j|>1.\end{aligned}$$ Let $ \U^+$, $ \U^0 $ and $ \U^-$ be the $ \Qq$-subalgebra of $ \U$ generated by $ E_{\alpha_i}$, $ K^{\pm 1}_{\alpha_i}$, and $ F_{\alpha_i}$ respectively, for $ i \in \I$. We introduce the divided exponent $ F^{(a)}_{\alpha_i } = F^a_{\alpha_i}/[a]!$, where $ a \ge 0 $, $ [ a ] = (q^{a}- q^{-a})/(q - q^{-1})$ and $ [ a ]! = [ 1][2 ] \cdots [ a]$. lease $ \mA = \Z[q, q^{-1}]$. Let $ _ \mA\U^{+}$ be the $ \mA$-subalgebra of $ \U^{+}$ generated by $ E^{(a)}_{\alpha_i}$ for various $ a \ge 0 $ and $ i \in \I$. Similarly get $ _ \mA\U^{-}$ be

} F_{\appha_j} -F_{\alpha_j} E_{\alpha_i} &= \dtlta_{i,j} \frac{K_{\alpha_n} -K^{-1}_{\alphe_i}}{q-q^{-1}}, \\ \djsplaybrdak[0] E_{\alpha_i}^2 E_{\alpha_j} +E_{\alpha_j} E_{\qlpha_u}^2 &= (q+q^{-1}) E_{\alpha_i} E_{\alpha_j} E_{\alpha_i}, \quad &&\tezt{if } |i-j|=1, \\ E_{\alpiz_i} E_{\alpmc_j} &= E_{\zppha_l} X_{\alpha_i}, \,\qquad\qauad &&\text{ix } |i-j|>1, \\ F_{\alpha_h}^2 W_{\appha_j} +F_{\alpha_j} F_{\alpha_i}^2 &= (q+q^{-1}) F_{\alpha_i} S_{\alpha_j} F_{\wlpha_i}, \quad\, &&\tgxt{if } |i-j|=1,\\ F_{\alpha_i} F_{\alpha_j} &= F_{\alpha_j} F_{\alpha_i}, \qquad\ \xquad &&\text{if } |i-j|>1.\end{aligned}$$ Let $\U^+$, $\U^0$ and $\U^-$ be hhe $\Qq$-subalgebra ov $\U$ generatgs br $E_{\alpha_i}$, $K^{\pm 1}_{\alpha_i}$, and $F_{\alpha_i}$ reapectively, for $i \in \I$. We introdjce tke divided koseg $F^{(a)}_{\alpha_i} = H^a_{\alphw_i}/[a]!$, where $a \nv 0$, $[a] = (q^{d}- q^{-a})/(q-q^{-1})$ snd $[a]! = [1][2] \cdots [s]$. Lxt $\mQ =\Z[q,q^{-1}]$. Let $_\mA\U^{+}$ be the $\kA$-subalgebra of $\U^{+}$ denerated bv $E^{(a)}_{\alpha_i}$ for variouw $q \ge 0$ and $i \iv \I$. Rimjlerlg let $_\lA\U^{-}$ be

} F_{\alpha_j} -F_{\alpha_j} E_{\alpha_i} &= \delta_{i,j} \frac{K_{\alpha_i} \displaybreak[0] E_{\alpha_j} +E_{\alpha_j} &= (q+q^{-1}) E_{\alpha_i} |i-j|=1, E_{\alpha_i} E_{\alpha_j} &= E_{\alpha_i}, \,\qquad\qquad &&\text{if |i-j|>1, \\ F_{\alpha_i}^2 F_{\alpha_j} +F_{\alpha_j} F_{\alpha_i}^2 (q+q^{-1}) F_{\alpha_i} F_{\alpha_j} F_{\alpha_i}, \quad\, &&\text{if } |i-j|=1,\\ F_{\alpha_i} F_{\alpha_j} &= F_{\alpha_j} F_{\alpha_i}, \qquad &&\text{if } |i-j|>1.\end{aligned}$$ Let $\U^+$, $\U^0$ and $\U^-$ be the $\Qq$-subalgebra of generated $E_{\alpha_i}$, 1}_{\alpha_i}$, $F_{\alpha_i}$ respectively, for $i \in \I$. We introduce the divided power $F^{(a)}_{\alpha_i} = F^a_{\alpha_i}/[a]!$, where $a 0$, $[a] = (q^{a}- q^{-a})/(q-q^{-1})$ and $[a]! = \cdots [a]$. Let $\mA Let $_\mA\U^{+}$ be the $\mA$-subalgebra $\U^{+}$ by $E^{(a)}_{\alpha_i}$ various \ge and $i \in Similarly let $_\mA\U^{-}$ be

} F_{\alpha_j} -F_{\alpha_j} E_{\alpha_i} &= \deltA_{i,j} \frac{K_{\alPha_i} -K^{-1}_{\AlpHa_i}}{Q-q^{-1}}, \\ \DispLaybReak[0] E_{\alpha_i}^2 E_{\alPHa_j} +E_{\Alpha_j} E_{\alpha_i}^2 &= (q+q^{-1}) E_{\alpha_i} e_{\alphA_j} e_{\AlphA_I}, \qUad &&\teXt{if } |i-j|=1, \\ E_{\ALpHA_I} E_{\aLpHa_J} &= E_{\aLpHA_j} e_{\alphA_i}, \,\qQuad\qquAd &&\text{if } |i-j|>1, \\ f_{\alPhA_i}^2 F_{\alpha_j} +F_{\alPHa_J} F_{\alpha_i}^2 &= (q+q^{-1}) f_{\alPha_i} F_{\alpha_j} F_{\AlpHa_i}, \quaD\, &&\tExt{IF } |i-j|=1,\\ F_{\aLphA_i} F_{\alPha_j} &= F_{\aLPha_j} F_{\aLpha_i}, \qquaD\ \qQUad &&\texT{If } |i-j|>1.\end{ALIgNed}$$ LEt $\U^+$, $\U^0$ and $\U^-$ be the $\Qq$-suBAlGEbra of $\U$ generatEd by $E_{\aLpHA_i}$, $k^{\PM 1}_{\alPha_I}$, and $F_{\alpha_I}$ rEspecTIvely, foR $I \iN \i$. wE inTRoduce the diviDed power $F^{(a)}_{\aLPha_I} = F^a_{\alpHa_I}/[a]!$, wHEre $a \ge 0$, $[A] = (q^{a}- q^{-a})/(Q-q^{-1})$ ANd $[a]! = [1][2] \Cdots [a]$. Let $\mA =\z[q,q^{-1}]$. LEt $_\mA\U^{+}$ be thE $\mA$-subALgebra oF $\u^{+}$ generaTed by $E^{(A)}_{\alPha_I}$ for VArIoUs $a \Ge 0$ ANd $i \IN \I$. simILarLy let $_\mA\U^{-}$ Be

} F_{\alpha_j} -F_{\alpha_ j} E_{\alp ha_i} &= \d el ta_{ i,j} \frac{K_{\alp h a_i} -K^{-1}_{\alpha_i}}{ q-q^{ -1 } }, \ \ \ displ aybreak [ 0] E_ {\ al pha _i } ^2 E_{\ alp ha_j} + E_{\alpha_ j}E_ {\alpha_i}^2 &= (q+q^{- 1}) E_{\alpha_i } E _{\alp ha _j} E_{\a lph a_i}, \qua d &&\te xt{if } | i- j |=1, \ \ E_{\ a l ph a_i} E_{\alpha_j} &=E _{ \ alpha_j} E_{\a lpha_i }, \ , \ qqu ad\ qquad &&\t ex t{if} |i-j|> 1 ,\ \ F _ {\alpha_i}^2F_{\alpha_j } +F _{\alp ha _j} F_{\al pha_i }^ 2 &= (q+q^{-1 }) F _{\alpha_ i} F_{ \ alpha_j } F_{\al pha_i} , \qu ad\, && \t ext {i f }| i- j|= 1 ,\\ F_{\a lp ha _i} F _{\a l p h a _j}&=F_{\ alpha _j} F_{\alpha _i} , \ q qua d\ \q quad&&\t ex t{if} |i-j |>1.\ en d{aligned}$$ L et $ \U^+$, $\ U^0 $and $ \U^-$ be the $\ Qq$ -subalg ebra of $\U $g e n er ated by $E_{\alpha _i } $ ,$K^{\pm1}_{\a l ph a_ i }$, and$F _{\ alph a _ i}$ r espe c ti vely, fo r $i \ i n\I $. We i nt roduce t hediv idedp ower $F^{( a)}_{\al pha_i } = F^a_{\alpha _ i}/[a]!$, whe r e$ a \ g e 0$ , $ [a] = (q^{a }- q ^ {-a} )/(q - q^ {-1 } )$ an d $[a ]! =[ 1][2] \cdots [a]$.Le t $\mA =\Z[ q,q^{-1}]$. L et $_\mA\U ^ { + }$ be th e $\ m A$ - subalgebra of$\U^{ +}$ genera t ed by $E ^{(a) }_{\alph a_i}$ for v arious $ a \ ge0$and $ i\in \I$. Simi l a rlyle t $_\mA \U^ {-}$ be

} F_{\alpha_j}_-F_{\alpha_j} E_{\alpha_i}_&= \delta_{i,j} \frac{K_{\alpha_i} -K^{-1}_{\alpha_i}}{q-q^{-1}},_\\ \displaybreak[0] __E_{\alpha_i}^2 E_{\alpha_j}_+E_{\alpha_j}_E_{\alpha_i}^2 &=_(q+q^{-1}) E_{\alpha_i} E_{\alpha_j}_E_{\alpha_i}, \quad &&\text{if_} |i-j|=1, \\ __E_{\alpha_i} E_{\alpha_j} &= E_{\alpha_j} E_{\alpha_i}, \,\qquad\qquad &&\text{if } |i-j|>1, \\ F_{\alpha_i}^2 F_{\alpha_j}_+F_{\alpha_j}_F_{\alpha_i}^2 _&=_(q+q^{-1})_F_{\alpha_i} F_{\alpha_j} F_{\alpha_i}, \quad\,_&&\text{if } |i-j|=1,\\ F_{\alpha_i}_F_{\alpha_j} &=_F_{\alpha_j} F_{\alpha_i}, \qquad\ \qquad &&\text{if } |i-j|>1.\end{aligned}$$ Let_$\U^+$,_$\U^0$ and $\U^-$_be the $\Qq$-subalgebra of $\U$ generated by $E_{\alpha_i}$, $K^{\pm_1}_{\alpha_i}$, and $F_{\alpha_i}$ respectively, for $i_\in \I$. We_introduce_the_divided power $F^{(a)}_{\alpha_i} =_F^a_{\alpha_i}/[a]!$, where $a \ge 0$, $[a]_= (q^{a}- q^{-a})/(q-q^{-1})$ and $[a]! =_[1][2] \cdots [a]$. Let $\mA =\Z[q,q^{-1}]$. Let_$_\mA\U^{+}$ be the $\mA$-subalgebra of $\U^{+}$_generated by $E^{(a)}_{\alpha_i}$ for various_$a \ge_0$ and $i \in \I$._Similarly let $_\mA\U^{-}$_be

base field (cf. after). One of the aims of this note is to show that, for function fields in positive characteristic, even a weak form of this is false. Usually, when one wants to prove a theorem on the arithmetic of rational (algebraic) points of varieties over global fields, the situation is more favorable in the function fields case. This is principally due to the fact that, over these fields, an horizontal derivation is available (there is a non trivial derivation over the base field). This is why many statements which are conjectural over varieties over number fields are proved in the analogous situation over function fields. Consequently, it is widely believed that a conjecture in this theory should be checked before over function fields and then, once the proof is well understood there, one should try to attack it for varieties over number fields. We want to show, mainly by examples, that some part of height theory seems to better behave over number fields then over function fields. This, again, is due to the existence of the so called isotrivial varieties (the analogous of which do not seem to exist over number fields). In the last part of this paper we will construct explicit examples of surfaces over a function field of positive characteristic which are of general type, are not birational to isotrivial surfaces and which are dominated by a surface defined over the base field. These surfaces will provide counterexamples to statements similar to Lang and Vojta conjectures. The fact that part the analogy is broken by the existence of isotrivial varieties is, in our opinion, a very important issue which should be analyzed more deeply. A better comprehension of it would probably improve aspects of the analogy and will lead to a development of the common language. This will allow to perhaps better formulate the leading conjectures of the theory. This note is based on the talk I gave at the “Terzo incontro italiano di teoria dei numeri” held in Pisa in september 2015. I would warmly thank the organizers, in particolar Andrea Bandini and Ilaria Del Corso for the perfect organization of the meeting and for the possibility they gave to me to give a talk. Notations, terminology ====================== In the sequel $K$ will be a global field. Thus $K$ may be either a number field or the field of rational function of a smooth projective curve $B$ over the complex numbers or the field of rational functions of a smooth projective curve over an algebraically closed field

base field (cf. after). One of the aim of this bill is to show that, for function fields in plus characteristic, even a weak shape of this is fake. Usually, when one wants to prove a theorem on the arithmetic of rational (algebraic) compass point of varieties over ball-shaped fields, the situation is more favorable in the function fields case. This is principally ascribable to the fact that, over these fields, an horizontal derivation is available (there is a non superficial derivation over the base plain). This is why many statements which are conjectural over assortment over number fields are proved in the analogous situation over function field. Consequently, it is widely believed that a conjecture in this theory should be check before over function fields and then, once the proof is well understood there, one should try to attack it for varieties over number fields. We desire to prove, mainly by exemplar, that some character of height theory seems to better behave over number fields then over affair fields. This, again, is due to the existence of the so called isotrivial varieties (the analogous of which do not appear to exist over number fields). In the last part of this newspaper we will construct denotative example of surfaces over a function field of positive characteristic which are of general type, are not birational to isotrivial surface and which are dominated by a surface defined over the base playing field. These surfaces will put up counterexamples to statements alike to Lang and Vojta conjectures. The fact that region the analogy is violate by the existence of isotrivial variety is, in our opinion, a very important offspring which should be analyzed more profoundly. A better comprehension of it would probably better aspects of the analogy and will lead to a development of the common speech. This will allow to perhaps well formulate the leading conjectures of the theory. This note is based on the lecture I gave at the “ Terzo incontro italiano di teoria dei numeri ” held in Pisa in september 2015. I would warmly thank the organizers, in particolar Andrea Bandini and Ilaria Del Corso for the perfect organization of the meeting and for the possibility they hold to me to give a talk. Notations, terminology = = = = = = = = = = = = = = = = = = = = = = In the sequel $ K$ will be a global field. Thus $ K$ may be either a number battlefield or the field of rational function of a placid projective curve $ B$ over the complex numbers or the field of rational functions of a placid projective curve over an algebraically closed field

bade field (cf. after). One of the aims of thnw note is to show thxt, for function fields in podirive xharacteristic, even a deak form of this is halse. Usually, whei one wakcs to irove e theorem on thg arithmetic of rational (angdbxaic) points of varieties over global fields, tje situation if moge favkganle in the function fields case. Fhis is principally cue to the fact that, over hhesf fields, an horizojtal derivajjon us available (there is a non triviam derivation over the base fiela). Thix is why mqnt shdtements whmch arv conjectural over vasieties over number flelds arw proved in the analojous situation over sunction xizlds. Consequently, it us wideny baliexwd ghau a cknjectkre in this tgeory shoule be checked before odvt function fjelds wnq then, once the proof is well understoog tgere, one should try to qttack it for varietigs over nuiber fields. We want to show, mainly by examples, thdt sojd pcvb of yelght theory seems to better behave over numbew foekds then over nunction fields. Thos, abwin, is due to the erjsfence of the so capled isjtrivual varieuies (yhe analogous of which do nit seem to eqist over number fieldd). In the lasc part of tnis paper we will constxuct esplicit exalples of ajrfaces over a fjncnion field of positive charactqristic wiich cre of gdnersl typq, are not hiratlmnal to isotrivial surfcces dnd which wre dominated by a surface defiixd over the bssa fpeld. Thesz surfsces will prodide counterexcmples tj stagements sijilar tm Lang and Dojta conjectgtes. The fact tiat part ehe qnaligy is cfoken by the ecistence of isotrivual varieties is, ik our kpinion, a very nokoetant issue whoch shjupd bq analyzed mose ddepuu. A bdtter comprchevsiom of it would probabny ijprove aspects of yhc analogy and wilj lead to a drvelopment of the fommoi langnage. Tnis will allow to perhaps better rormulate thc leading contectmres of the thzory. This note is based on the talk I gavx at the “Terzo incontro italiano di teoria cei numeri” ield ig Pisa in september 2015. I would qarmly thank the prganizers, in particomar Angrea Handini and Ilaria Del Corso for the perfect organization of the meeting qnd foc ehe possibimity they gcve to me to jite a talk. Notationx, terminology ====================== In the sequel $K$ wiul be a cljbal field. Thus $K$ may be eitnef a number figld or the field of rationzl functoon of a smooth projective curvr $B$ over the complex numners or tye fisld of ratmonal functions of a smootk projwctive xurvc over an algecraocwlly closfd fmelb

base field (cf. after). One of the this is to that, for function a form of this false. Usually, when wants to prove a theorem on arithmetic of rational (algebraic) points of varieties over global fields, the situation is favorable in the function fields case. This is principally due to the fact over fields, horizontal is available (there is a non trivial derivation over the base field). This is why many which are conjectural over varieties over number fields proved in the analogous over function fields. Consequently, it widely that a in theory be checked before function fields and then, once the proof is well understood there, one should try to attack it varieties over We want show, by that some part theory seems to better behave over over function fields. This, again, is due to existence of so called isotrivial varieties (the analogous which do not seem to exist over number In the last part of this paper we will construct explicit examples of surfaces over field of positive characteristic are of general are birational isotrivial and which dominated by a surface defined over the base field. These surfaces provide counterexamples to statements similar to Lang and Vojta conjectures. that the analogy is by the existence of varieties in our opinion, a issue should deeply. better of it would probably aspects of the analogy and lead to a development will allow to perhaps better formulate the leading of the theory. This note is based the talk I gave at the “Terzo incontro italiano di teoria dei held in september 2015. I would warmly thank the organizers, particolar Andrea Bandini and Del Corso for the perfect organization of the meeting for possibility they to me to a talk. Notations, ====================== In the will be global Thus either a number field or the of function of a smooth projective $B$ complex numbers or the field of rational functions of a smooth projective curve over an field

base field (cf. after). One of the aIms of this nOte is To sHow ThAt, foR funCtion fields in pOSitiVe characteristic, even a wEak foRm OF thiS Is False. usually, WHeN ONe wAnTs To pRoVE a TheorEm oN the ariThmetic of rAtiOnAl (algebraic) pOInTs of varietIes Over global fiEldS, the siTuAtiON is moRe fAvoraBle in tHE functIon fields CaSE. This iS PrincipALLy Due tO the fact that, over tHEsE Fields, an horizoNtal deRiVAtION is AvaIlable (therE iS a non TRivial dERiVATIon OVer the base fieLd). This is why MAny StatemEnTs wHIch are ConjeCtURal Over varietiEs ovEr number fIelds aRE proved IN the anaLogous SitUatIon oVEr FuNctIoN FieLDs. conSEquEntly, it iS wIdEly beLievED THAt a cOnjEctuRe in tHis theory shouLd bE cheCKed BeforE over FuncTiOn fieLds and Then, oNcE the proof is well UndeRstood theRe, oNe ShoUlD try tO Attack It fOr vArietieS over nuMBer FiELDS. WE want to show, mainly bY eXAMpLes, that sOme parT Of HeIGht theorY sEemS to bETTer beHave OVeR number fIelds tHEn OvEr functIoN fieldS. THis, AgaIn, is dUE to tHe exisTence of tHe so cALled isotrivial VArieties (the anALoGOUs OF whiCh dO not seem to eXist OVer nUmbeR FiEldS). in the Last pArT Of THis paper we will constRuCt explIcit eXamples of surfAces over a fUNCTion fielD of pOSiTIve characterisTic whIch are of geNEral type, Are noT biratioNal to isotRIVial surfAceS anD whIch ARE dOminated by a suRFAce dEfIned oveR thE base fiEld. theSe sUrfAcEs will proVide counTeReXaMpLes To staTEments siMiLar To lanG and VOJta conJectuRes. THe FaCT thAt part tHE aNALogy Is BrOken By tHe ExistEnce OF isOtriviaL varietieS is, IN our OpInIon, a verY important issUe Which shoulD bE anAlyzed MORe deeply. a better comprehension of iT Would prObaBly imProvE aspects oF thE analoGy aND will lEad to a DevelOpMenT OF the cOMMoN laNgUage. This wiLL AllOw to pErHaps Better fOrmulate the leading COnjEctures of the tHeoRy. ThIS NoTe iS BaSEd oN tHE taLK i gave at the “Terzo Incontro itAlIAnO di teoria dEI nuMeRi” held iN Pisa in SepteMBer 2015. I wouLd warmly tHank the orGaNizeRS, In pArticolar ANdrea BanDini and IlARia DeL coRso foR thE perfeCt OrgAnizaTion of THe mEetinG and foR tHe possIbiliTy They gave To me to give a talk. NotationS, termiNologY ====================== In The sequel $k$ wiLL be A global fiEld. THus $K$ may be eIthEr a NumbeR fiELd or tHe fiELd Of rATionaL funCTion of a smOOtH prOJEcTive curve $B$ oVER The ComplEx nUMbers oR the Field of rational fuNCtions of a smootH proJECtiVe cURve oVeR an algebraicalLy cLoSED field

base field (cf. after). O ne of theaimsofthi snote isto show that,f or f unction fields in posi tivech a ract e ri stic, even a we a k fo rm o f t hi s i s fal se. Usual ly, when o newa nts to prove atheorem on th e arithmetic of ratio na l ( a lgebr aic ) poi nts of variet ies overgl o bal fi e lds, th e si tuat ion is more favor a bl e in the functi on fie ld s c a s e.Thi s is princ ip allyd ue to t h ef a c t t h at, over thes e fields, a n ho rizont al de r ivatio n isav a ila ble (thereis a non triv ial de r ivation over th e base fi eld ). T h is i s w hy man y s tat e men ts which a re conj ectu r a l over va riet ies o ver number fi eld s ar e pr ovedin th e an al ogous situa tionov er function fie lds. Conseque ntl y, it i s wid e ly bel iev edthat aconject u rein t h is theory should bech e c ke d before overf un ct i on field sand the n , once the pr oof is w ell un d er st ood the re , onesh oul d t ry to atta ck itfor vari eties over number fi e lds. We wantt os h ow , mai nly by example s, t h at s omep ar t o f heig ht th eo r ys eems to better beha ve overnumbe r fields then over func t i o n fields . Th i s, again, is dueto th e existenc e of theso ca lled iso trivial v a r ieties ( the an alo gou s of which do not s eemto existove r numbe r f iel ds) . In the last part of t hi spa per we w i ll const ru ctex pli cit e x amples of s urfa ce so ver a func t io n fiel dof pos iti ve char acte r ist ic whic h are ofgen e ralty pe , are n ot birational t o isotrivi al su rfaces a nd which are dominated by a sur f ace def ine d ove r th e base fi eld . Thes e s u rfaces willprovi de co u n terex a m pl esto statement s sim ilarto Lan g and V ojta conjectures.The fact that pa rtthea n al ogy is bro ke n by t he existence of isotrivia lv ar ieties is, inou r opini on, a v ery i m portant issue wh ich shoul dbe a n a lyz ed more de eply. Abetter co m prehe n si on of it would p rob ablyimprov e as pects of th eanalog y and w ill lead to a development of th e comm on la ngu age. This wi l l a llow to p erha ps betterfor mul ate t hel eadin g co n je ctu r es of the theory. T hi s n o t eis based on t h e t alk I ga v e at t he “ Terzo incontro it a liano di teori a de i num eri ” hel din Pisa in sep tem be r 2015. Iwo uld warmlythank th eo rgani zers,in par ticolar A nd r ea Ban dini an d IlariaDel C o rso for t he perfec t or ga nizati on oft he m e e ting and for the poss i b ility the y gav eto me t o giv e a talk. Notations, termi nolo gy == ======= == ====== === == In the s e quel $K$willbe a gl ob al f iel d. Thu s $K $ may b e ei th era numberf i el d o rt hefiel d ofra tion al functi o n of a s moo t h proje ct ive c urve $ B $o v er the com ple x num b e rs or thef ield of ratio nal fu nction s of as moo th projec tiv e curve ove r an alge b rai ca llyclosed f ie ld

base_field (cf._after). One of the_aims of_this_note is_to_show that, for_function fields in_positive characteristic, even a_weak form of_this_is false. Usually, when one wants to prove a theorem on the arithmetic of rational_(algebraic)_points of_varieties_over_global fields, the situation is_more favorable in the function_fields case._This is principally due to the fact that,_over_these fields, an_horizontal derivation is available (there is a non trivial_derivation over the base field). This_is why many_statements_which_are conjectural over varieties_over number fields are proved in_the analogous situation over function fields._Consequently, it is widely believed that a_conjecture in this theory should be_checked before over function fields_and then,_once the proof is well_understood there, one_should try_to attack it_for varieties over number fields. We_want to show,_mainly by examples, that some part_of_height theory seems_to_better_behave over_number fields then_over_function fields._This,_again, is due to the existence_of_the so called isotrivial varieties (the analogous_of which do not_seem_to exist over number_fields). In the last part of_this paper we will construct explicit_examples of_surfaces over_a function field of positive characteristic which are of general type,_are not birational to isotrivial surfaces_and which are dominated_by a_surface_defined over the_base_field. These_surfaces will provide counterexamples to statements similar_to Lang_and Vojta conjectures. The fact that part_the analogy is broken_by_the existence of isotrivial varieties is,_in our opinion, a very important_issue which should be analyzed_more_deeply._A better comprehension of it_would probably improve aspects of the_analogy and will_lead to a development of the common_language._This will allow to perhaps better_formulate_the leading conjectures of the theory. This_note_is_based on the talk I_gave at the “Terzo incontro italiano_di teoria dei numeri” held in Pisa in september_2015. I would_warmly thank the organizers, in_particolar_Andrea_Bandini and Ilaria Del Corso for the perfect organization of_the meeting_and for the_possibility they gave to me to give a talk. Notations, terminology ====================== In_the sequel $K$ will be a global_field. Thus $K$ may be either a number field or the_field of rational function of a smooth projective_curve $B$ over the complex numbers or_the field of rational_functions_of a smooth projective curve over an_algebraically_closed_field

X_\gamma = \overline{{\textrm{span}}}^{\Vert \: \Vert_2} \{ \gamma_{m,n} \mid m,n \in \mathbb{N}_0 \}\subset \ell^2(W).$$ The following Lemma \[Lem=Blackbox\] collects all computational results we need further. As all the (rather easy) arguments are basically contained in [@Radulescu Lemma 1] we merely sketch the proof; all other proofs we give in this section will then be self-contained. \[Lem=Blackbox\] We have the following: 1. \[Item=BB1\] For $\gamma \in {{\mathbb{C}}_q^l[W]}, l \geq 1, m \geq 1, n \geq 0$, we have $$h_1 \gamma_{m,n} = \gamma_{m+1,n} + p \gamma_{m,n} + (L-1) \gamma_{m-1, n}.$$ 2. \[Item=BB2\] For $\gamma \in {{\mathbb{C}}_q^l[W]}\ominus S_l, l \geq 2, m \geq 0, n \geq 0$, we have $$h_1 \gamma_{m,n} = \gamma_{m+1,n} + p \gamma_{m,n} + (L-1) \gamma_{m-1, n}.$$ (Note that only the case $m=0$ was not already covered by ). 3. \[Item=BB3\] For $\beta \in {{\mathbb{C}}_q^1[W]}\ominus S_1, n \geq 0$, we have $$h_1 \beta_{0,n} = \beta_{1,n} + p \beta_{0,n} - \beta_{0, n-1}.$$ 4. \[Item=BB4\] \[Item=BlackboxOne\] For $\gamma \in {{\mathbb{C}}_q^l[W]}, l \geq 1$ we have, $$\begin{split} & q_{l+m+n+1}( h_1 h_m \gamma h_n) = q_{l+n+m+1}( h_1 q_{l+m+n}(h_m \gamma h_n)), \qquad \:m

X_\gamma = \overline{{\textrm{span}}}^{\Vert \: \Vert_2 } \ { \gamma_{m, n } \mid m, n \in \mathbb{N}_0 \}\subset \ell^2(W).$$ The following Lemma \[Lem = Blackbox\ ] collects all computational results we need far. As all the (preferably easy) arguments are basically contained in [ @Radulescu Lemma 1 ] we merely sketch the validation; all other proofs we give in this part will then be self - contained. \[Lem = Blackbox\ ] We have the pursuit: 1. \[Item = BB1\ ] For $ \gamma \in { { \mathbb{C}}_q^l[W ] }, lambert \geq 1, m \geq 1, n \geq 0 $, we have $ $ h_1 \gamma_{m, n } = \gamma_{m+1,n } + phosphorus \gamma_{m, n } + (L-1) \gamma_{m-1, n}.$$ 2. \[Item = BB2\ ] For $ \gamma \in { { \mathbb{C}}_q^l[W]}\ominus S_l, l \geq 2, m \geq 0, n \geq 0 $, we have $ $ h_1 \gamma_{m, n } = \gamma_{m+1,n } + p \gamma_{m, n } + (L-1) \gamma_{m-1, n}.$$ (Note that merely the case $ m=0 $ was not already covered by). 3. \[Item = BB3\ ] For $ \beta \in { { \mathbb{C}}_q^1[W]}\ominus S_1, n \geq 0 $, we have $ $ h_1 \beta_{0,n } = \beta_{1,n } + p \beta_{0,n } - \beta_{0, n-1}.$$ 4. \[Item = BB4\ ] \[Item = BlackboxOne\ ] For $ \gamma \in { { \mathbb{C}}_q^l[W ] }, liter \geq 1 $ we have, $ $ \begin{split } & q_{l+m+n+1 } (h_1 h_m \gamma h_n) = q_{l+n+m+1 } (h_1 q_{l+m+n}(h_m \gamma h_n) ), \qquad \:m

X_\galma = \overline{{\textrm{span}}}^{\Vtrt \: \Vert_2} \{ \gamma_{m,n} \mid k,n \in \jathbb{N}_0 \}\rubset \ell^2(W).$$ The following Lelmq \[Lem=Vlackbox\] collects all zomputatilnal resylts qe need fucfher. As all tgc (ratkec easy) argumentx are basiwally containeg kn [@Radulescu Lemma 1] we merely sketch ehe propf; all other projfs eq gibv ln this section will then be selr-contaiied. \[Lem=Blackbox\] Ee have the following: 1. \[Itel=BB1\] Vor $\gamma \in {{\mathbh{C}}_q^l[W]}, l \geq 1, m \dwq 1, n \geq 0$, wd have $$h_1 \gamma_{m,n} = \gammz_{m+1,n} + p \gamma_{m,n} + (L-1) \gamma_{m-1, n}.$$ 2. \[Iteo=BB2\] Fpr $\gamma \ib {{\natjtb{C}}_q^l[W]}\ominuw S_l, j \geq 2, m \geq 0, n \geq 0$, we havr $$h_1 \gamma_{m,n} = \gsmme_{m+1,n} + p \gamma_{m,n} + (L-1) \gamma_{m-1, i}.$$ (Note that only the case $m=0$ wds not already coveeee by ). 3. \[Itam=BB3\] Dor $\beua \mn {{\jathbb{F}}_q^1[W]}\kminus S_1, n \geq 0$, we have $$h_1 \beta_{0,n} = \beta_{1,m} + i \beta_{0,n} - \befa_{0, n-1}.$$ 4. \[Ytqm=BB4\] \[Item=BlackboxOne\] For $\gamma \in {{\mathbt{C}}_q^m[W]}, l \geq 1$ we have, $$\begin{wplit} & q_{l+m+n+1}( h_1 h_m \hamma h_n) = q_{l+n+m+1}( h_1 q_{l+m+n}(h_m \gamma h_n)), \qquad \:m

X_\gamma = \overline{{\textrm{span}}}^{\Vert \: \Vert_2} \{ \gamma_{m,n} \in \}\subset \ell^2(W).$$ following Lemma \[Lem=Blackbox\] need As all the easy) arguments are contained in [@Radulescu Lemma 1] we sketch the proof; all other proofs we give in this section will then self-contained. \[Lem=Blackbox\] We have the following: 1. \[Item=BB1\] For $\gamma \in {{\mathbb{C}}_q^l[W]}, l 1, \geq n 0$, we have $$h_1 \gamma_{m,n} = \gamma_{m+1,n} + p \gamma_{m,n} + (L-1) \gamma_{m-1, n}.$$ 2. \[Item=BB2\] $\gamma \in {{\mathbb{C}}_q^l[W]}\ominus S_l, l \geq 2, m 0, n \geq 0$, have $$h_1 \gamma_{m,n} = \gamma_{m+1,n} p + (L-1) n}.$$ that the case $m=0$ not already covered by ). 3. \[Item=BB3\] For $\beta \in {{\mathbb{C}}_q^1[W]}\ominus S_1, n \geq 0$, we have \beta_{0,n} = p \beta_{0,n} \beta_{0, 4. \[Item=BlackboxOne\] For $\gamma l \geq 1$ we have, $$\begin{split} h_m \gamma h_n) = q_{l+n+m+1}( h_1 q_{l+m+n}(h_m \gamma \qquad \:m

X_\gamma = \overline{{\textrm{span}}}^{\VErt \: \Vert_2} \{ \gamMa_{m,n} \mId m,N \in \MaThbb{n}_0 \}\subSet \ell^2(W).$$ The follOWing lemma \[Lem=Blackbox\] collecTs all CoMPutaTIoNal reSults we NEeD FUrtHeR. AS alL tHE (rAther EasY) argumeNts are basiCalLy Contained in [@RADuLescu Lemma 1] We mErely sketch tHe pRoof; alL oTheR ProofS we Give iN this sECtion wIll then be SeLF-contaINed. \[Lem=BLACkBox\] WE have the following: 1. \[iTeM=bB1\] For $\gamma \in {{\maThbb{C}}_q^L[W]}, L \GeQ 1, M \Geq 1, N \geQ 0$, we have $$h_1 \gaMmA_{m,n} = \gaMMa_{m+1,n} + p \gaMMa_{M,N} + (l-1) \GamMA_{m-1, n}.$$ 2. \[Item=BB2\] For $\gAmma \in {{\mathbB{c}}_q^l[w]}\ominuS S_L, l \gEQ 2, m \geq 0, n \Geq 0$, we HaVE $$h_1 \gAmma_{m,n} = \gamma_{M+1,n} + p \gAmma_{m,n} + (L-1) \gaMma_{m-1, n}.$$ (NOTe that oNLy the caSe $m=0$ was Not AlrEady COvErEd bY ). 3. \[ITEm=Bb3\] foR $\beTA \in {{\Mathbb{C}}_q^1[w]}\oMiNus S_1, n \Geq 0$, wE HAVE $$h_1 \beTa_{0,n} = \Beta_{1,N} + p \betA_{0,n} - \beta_{0, n-1}.$$ 4. \[Item=BB4\] \[iteM=BlaCKboXOne\] FOr $\gamMa \in {{\MaThbb{C}}_Q^l[W]}, l \geQ 1$ we haVe, $$\Begin{split} & q_{l+m+n+1}( h_1 H_m \gaMma h_n) = q_{l+n+m+1}( H_1 q_{l+M+n}(H_m \gAmMa h_n)), \qQUad \:m

X_\gamma = \overline{{\tex trm{span}} }^{\V ert \: \ Vert _2}\{ \gamma_{m, n } \m id m,n \in \mathbb{N}_ 0 \}\ su b set\ el l^2(W ).$$ Th e f o l low in gLem ma \[ Lem=B lac kbox\]collects a llco mputationalr es ults we ne edfurther. Asall the ( ra the r easy ) a rgume nts ar e basic ally cont ai n ed in[ @Radule s c uLemm a 1] we merely sk e tc h the proof; al l othe rp ro o f s w e g ive in thi ssecti o n willt he n b e s e lf-contained. \[Lem=Bla c kbo x\] We h ave the fo llowi ng : 1 . \[Item=B B1\] For $\ga mma \i n {{\mat h bb{C}}_ q^l[W] },l \ geq1 ,m\ge q1 , n \g eq0 $,we have$$ h_ 1 \ga mma_ { m , n } =\ga mma_ {m+1, n} + p \gamma _{m ,n}+ (L -1) \ gamma _{m- 1, n}.$ $ 2. \[It em =BB2\] For $\ga mma\in {{\ma thb b{ C}} _q ^l[W] } \ominu s S _l, l \geq 2, m \ g eq0, n \g eq 0$, we have $$h _1 \ ga mma_{m,n } = \g a mm a_ { m+1,n} + p \g amma _ { m,n}+ (L - 1) \gamma_ {m-1,n }. $$ (Noteth at onl ythe ca se $m = 0$ w as not already cove r ed by ). 3. \ [Item=BB3\] F o r$ \ be t a \i n { {\mathbb{C} }_q^ 1 [W]} \omi n us S_ 1 , n \ge q0 $, we have $$h_1 \beta _{ 0,n} = \bet a_{1,n} + p \ beta_{0,n} - \beta_ {0,n -1 } .$$ 4. \[Ite m=BB4 \] \[Item= B lackboxO ne\]For $\ga mma \in { { \ mathbb{C }}_ q^l [W] },l \g eq 1$ we have , $$\b eg in{spli t} &q_{ l+m +n+ 1}( h _1 h_m \g amma h_n )=q_ {l +n+ m+1}( h_1 q_{ l+ m+n }( h_m \gam m a h_n) ), \q quad \ :m

X_\gamma =_\overline{{\textrm{span}}}^{\Vert \:_\Vert_2} \{ \gamma_{m,n}_\mid m,n_\in_\mathbb{N}_0 \}\subset_\ell^2(W).$$_The following Lemma_\[Lem=Blackbox\] collects all_computational results we need_further. As all_the_(rather easy) arguments are basically contained in [@Radulescu Lemma 1] we merely sketch the_proof;_all other_proofs_we_give in this section will_then be self-contained. \[Lem=Blackbox\] We have_the following: 1._ \[Item=BB1\] For $\gamma \in {{\mathbb{C}}_q^l[W]}, l \geq_1,_m \geq 1,_n \geq 0$, we have $$h_1 \gamma_{m,n} = \gamma_{m+1,n}_+ p \gamma_{m,n} + (L-1) \gamma_{m-1,_n}.$$ 2. \[Item=BB2\]_For_$\gamma_\in {{\mathbb{C}}_q^l[W]}\ominus S_l, l_\geq 2, m \geq 0, n_\geq 0$, we have $$h_1 \gamma_{m,n}_= \gamma_{m+1,n} + p \gamma_{m,n} + (L-1)_\gamma_{m-1, n}.$$ (Note that only the_case $m=0$ was not already_covered by_). 3. \[Item=BB3\] For $\beta_\in {{\mathbb{C}}_q^1[W]}\ominus S_1,_ _ n \geq_0$, we have $$h_1 \beta_{0,n} =_\beta_{1,n} + p_\beta_{0,n} - \beta_{0, n-1}.$$ 4.__\[Item=BB4\] \[Item=BlackboxOne\] For_$\gamma_\in_{{\mathbb{C}}_q^l[W]}, l_\geq 1$ we_have,_$$\begin{split} __ & q_{l+m+n+1}( h_1 h_m_\gamma_h_n) = q_{l+n+m+1}( h_1 q_{l+m+n}(h_m \gamma_h_n)), \qquad \:m

\, \displaystyle \sum_{P \in \S} \kappa_{P} \pgf^{(|P|-1)} \prod_{W \in P}{d_{W}}}.\end{aligned}$$ Evaluating the derivative we get $$\begin{aligned} \bigg\{ \frac{d^n}{d t^n} & t^{|P|-1} \, \pgf^{(|P|-1)}(t \gamma) \bigg\}_{t=0} \nonumber \\ &= \begin{cases} 0 & \text{if $n < |P|-1$} \\ \displaystyle \frac{n!}{(n-|P|+1)!} \pgf^{(n)}(0) \, \gamma^{n-|P|+1} & \text{if $n \geq |P|-1$}. \end{cases}\end{aligned}$$ We also have $\pgf^{(n)}(0) = n! \, p_{k+1|k}(n)$, hence $$\begin{aligned} & p_{k+1|k+1}(n) = p_{k+1|k}(n) \, \times \nonumber \\ & \quad \frac{\displaystyle \sum_{\substack{P \in \S \\ |P| \leq n+1}} \frac{n!}{(n-|P|+1)!} \left( \prod_{j=1}^{s} {C^{(m_{j}-|P|_{j})}_{j}(0)} \right) \, \gamma^{n-|P|+1} \prod_{W \in P}{d_{W}}} {\displaystyle \sum_{P \in \S} \kappa_{P} \pgf^{(|P|-1)} \prod_{W \in P}{d_{W}}}.\end{aligned}$$ We thus have $$\begin{aligned} \frac{p_{k+1|k+1}(n)}{p_{k+1|k}(n)} &= \frac{\displaystyle \sum_{\substack{P \in \S \\ |P| \leq n+1}} \left(\kappa_{P} \frac{n!}{(n-|P|+1)!} \gamma^{n-|P|+1} \prod_{W \in P}{d_{W}}\right)} {\displaystyle \sum

\, \displaystyle \sum_{P \in \S } \kappa_{P } \pgf^{(|P|-1) } \prod_{W \in P}{d_{W}}}.\end{aligned}$$ Evaluating the derivative we get $ $ \begin{aligned } \bigg\ { \frac{d^n}{d t^n } & t^{|P|-1 } \, \pgf^{(|P|-1)}(t \gamma) \bigg\}_{t=0 } \nonumber \\ & = \begin{cases } 0 & \text{if $ n < |P|-1 $ } \\ \displaystyle \frac{n!}{(n-|P|+1)! } \pgf^{(n)}(0) \, \gamma^{n-|P|+1 } & \text{if $ n \geq |P|-1 $ }. \end{cases}\end{aligned}$$ We besides take $ \pgf^{(n)}(0) = n! \, p_{k+1|k}(n)$, hence $ $ \begin{aligned } & p_{k+1|k+1}(n) = p_{k+1|k}(n) \, \times \nonumber \\ & \quad \frac{\displaystyle \sum_{\substack{P \in \S \\ |P| \leq n+1 } } \frac{n!}{(n-|P|+1)! } \left (\prod_{j=1}^{s } { C^{(m_{j}-|P|_{j})}_{j}(0) } \right) \, \gamma^{n-|P|+1 } \prod_{W \in P}{d_{W } } } { \displaystyle \sum_{P \in \S } \kappa_{P } \pgf^{(|P|-1) } \prod_{W \in P}{d_{W}}}.\end{aligned}$$ We thus have $ $ \begin{aligned } \frac{p_{k+1|k+1}(n)}{p_{k+1|k}(n) } & = \frac{\displaystyle \sum_{\substack{P \in \S \\ |P| \leq n+1 } } \left(\kappa_{P } \frac{n!}{(n-|P|+1)! } \gamma^{n-|P|+1 } \prod_{W \in P}{d_{W}}\right) } { \displaystyle \sum

\, \dlsplaystyle \sum_{P \in \S} \kakpa_{P} \pgf^{(|P|-1)} \prod_{W \iu P}{d_{W}}}.\env{alignes}$$ Evaluaging the derivative we get $$\bxgin{qligntb} \bigg\{ \frac{d^n}{d t^n} & t^{|P|-1} \, \pgf^{(|P|-1)}(t \gwmma) \bigt\}_{t=0} \nibumber \\ &= \bejjn{cases} 0 & \tzxv{if $n < |P|-1$} \\ \displaysdyle \frac{n!}{(n-|P|+1)!} \pcf^{(v)}(0) \, \gamma^{n-|P|+1} & \text{if $n \geq |P|-1$}. \end{casef}\end{alibnfd}$$ We also havg $\pgf^{(m)}(0) = n! \, i_{k+1|j}(n)$, hence $$\begin{aligned} & p_{k+1|k+1}(n) = p_{i+1|k}(n) \, \tikes \nonumber \\ & \quad \frac{\displaystyle \sum_{\dubshack{P \in \S \\ |P| \leq j+1}} \frac{n!}{(n-|P|+1)!} \lqdt( \prod_{j=1}^{s} {C^{(m_{g}-|P|_{j})}_{j}(0)} \right) \, \gamma^{n-|P|+1} \prkd_{W \in P}{d_{W}}} {\displaystyle \sum_{P \in \S} \kappa_{P} \pgd^{(|P|-1)} \prlg_{W \in P}{d_{W}}}.\env{alignvd}$$ We thus hayv $$\begin{dligned} \grac{p_{k+1|k+1}(n)}{p_{k+1|k}(n)} &= \nrac{\dmsplqystyle \sum_{\substack{P \mn \S \\ |P| \leq n+1}} \lest(\kappa_{P} \xrcc{n!}{(n-|P|+1)!} \gamma^{n-|P|+1} \prod_{W \ib P}{d_{W}}\richt)} {\dkwplxysuylx \shm

\, \displaystyle \sum_{P \in \S} \kappa_{P} \pgf^{(|P|-1)} P}{d_{W}}}.\end{aligned}$$ the derivative get $$\begin{aligned} \bigg\{ \pgf^{(|P|-1)}(t \bigg\}_{t=0} \nonumber \\ \begin{cases} 0 & $n < |P|-1$} \\ \displaystyle \frac{n!}{(n-|P|+1)!} \, \gamma^{n-|P|+1} & \text{if $n \geq |P|-1$}. \end{cases}\end{aligned}$$ We also have $\pgf^{(n)}(0) = \, p_{k+1|k}(n)$, hence $$\begin{aligned} & p_{k+1|k+1}(n) = p_{k+1|k}(n) \, \times \nonumber \\ & \frac{\displaystyle \in \\ \leq n+1}} \frac{n!}{(n-|P|+1)!} \left( \prod_{j=1}^{s} {C^{(m_{j}-|P|_{j})}_{j}(0)} \right) \, \gamma^{n-|P|+1} \prod_{W \in P}{d_{W}}} {\displaystyle \sum_{P \in \S} \pgf^{(|P|-1)} \prod_{W \in P}{d_{W}}}.\end{aligned}$$ We thus have $$\begin{aligned} &= \frac{\displaystyle \sum_{\substack{P \in \\ |P| \leq n+1}} \left(\kappa_{P} \gamma^{n-|P|+1} \in P}{d_{W}}\right)} \sum

\, \displaystyle \sum_{P \in \S} \kappa_{P} \Pgf^{(|P|-1)} \prod_{W \iN P}{d_{W}}}.\eNd{aLigNeD}$$ EvaLuatIng the derivatiVE we gEt $$\begin{aligned} \bigg\{ \frac{D^n}{d t^n} & T^{|P|-1} \, \PGf^{(|P|-1)}(t \GAmMa) \bigG\}_{t=0} \nonumBEr \\ &= \BEGin{CaSeS} 0 & \teXt{IF $n < |p|-1$} \\ \dispLayStyle \frAc{n!}{(n-|P|+1)!} \pgf^{(n)}(0) \, \gAmmA^{n-|p|+1} & \text{if $n \geq |P|-1$}. \ENd{Cases}\end{alIgnEd}$$ We also have $\Pgf^{(N)}(0) = n! \, p_{k+1|k}(n)$, HeNce $$\BEgin{aLigNed} & p_{k+1|K+1}(n) = p_{k+1|k}(n) \, \TImes \noNumber \\ & \quaD \fRAc{\dispLAystyle \SUM_{\sUbstAck{P \in \S \\ |P| \leq n+1}} \frac{n!}{(N-|p|+1)!} \lEFt( \prod_{j=1}^{s} {C^{(m_{j}-|P|_{j})}_{j}(0)} \Right) \, \gAmMA^{n-|p|+1} \PRod_{w \in p}{d_{W}}} {\displayStYle \suM_{p \in \S} \kapPA_{P} \PGF^{(|p|-1)} \prOD_{W \in P}{d_{W}}}.\end{aliGned}$$ We thus hAVe $$\bEgin{alIgNed} \FRac{p_{k+1|k+1}(N)}{p_{k+1|k}(n)} &= \FrAC{\diSplaystyle \sUm_{\suBstack{P \in \s \\ |P| \leq n+1}} \LEft(\kappA_{p} \frac{n!}{(n-|p|+1)!} \gamma^{N-|P|+1} \pRod_{w \in P}{D_{w}}\rIgHt)} {\dIsPLaySTyLe \sUM

\, \displaystyle \sum_{P\in \S} \k appa_ {P} \p gf ^{(| P|-1 )} \prod_{W \i n P}{ d_{W}}}.\end{aligned}$ $ Eva lu a ting th e der ivative we g et$$ \b egi n{ a li gned} \b igg\{ \ frac{d^n}{ d t ^n } & t^{|P|-1 } \ , \pgf^{(| P|- 1)}(t \gamma ) \ bigg\} _{ t=0 } \non umb er \\ &= \b e gin{ca ses} 0 & \ t ext{if$ n < |P| -1$} \\ \ d is p laystyle \frac {n!}{( n- | P| + 1 )!} \p gf^{(n)}(0 )\, \g a mma^{n- | P| + 1 } &\ text{if $n \g eq |P|-1$}. \end{ ca ses } \end{a ligne d} $ $ W e also have $\p gf^{(n)}( 0) = n ! \, p_{ k +1|k}(n )$, he nce $$ \beg i n{ al ign ed } &p _{ k+1 | k+1 }(n) = p _{ k+ 1|k}( n) \ , \ t imes \n onum ber \ \ & \quad \fr ac{ \dis p lay style \sum _{\s ub stack {P \in \S \ \|P| \leq n+1}}\fra c{n!}{(n- |P| +1 )!} \ l eft( \ pro d_{ j=1}^{s } {C^{( m _{j }- | P | _{ j})}_{j}(0)} \righ t) \ ,\gamma^{ n-|P|+ 1 }\p r od_{W \i nP}{ d_{W } } } {\d i sp laystyle \sum_ { P\i n \S} \ ka ppa_{P }\pg f^{ (|P|- 1 )} \ prod_{ W \in P} {d_{W } }}.\end{aligne d }$$ We thus h a ve $ $\ b egin {al igned} \fra c{p_ { k+1| k+1} ( n) }{p _ {k+1| k}(n) }& =\ frac{\displaystyle\s um_{\s ubsta ck{P \in \S \ \ |P| \leq n + 1}} \le f t( \ kappa_{P} \fra c{n!} {(n-|P|+1) ! } \gamma ^{n-| P|+1} \p rod_{W \i n P}{d_{W} }\r igh t)} {\ displaystyle\ s um

\,_\displaystyle \sum_{P_\in \S} \kappa_{P} \pgf^{(|P|-1)}_\prod_{W \in_P}{d_{W}}}.\end{aligned}$$_Evaluating the_derivative_we get $$\begin{aligned} \bigg\{_\frac{d^n}{d t^n} &_t^{|P|-1} \, \pgf^{(|P|-1)}(t \gamma)_\bigg\}_{t=0} \nonumber \\ &=_\begin{cases} _ 0 & \text{if $n < |P|-1$} \\ __ ___ \displaystyle \frac{n!}{(n-|P|+1)!} \pgf^{(n)}(0)_\, \gamma^{n-|P|+1} & \text{if $n_\geq |P|-1$}. _ \end{cases}\end{aligned}$$ We also have $\pgf^{(n)}(0)_=_n! \, p_{k+1|k}(n)$,_hence $$\begin{aligned} & p_{k+1|k+1}(n) = p_{k+1|k}(n) \, \times \nonumber \\ &_\quad \frac{\displaystyle \sum_{\substack{P \in \S \\_|P| \leq n+1}}_\frac{n!}{(n-|P|+1)!}_ _ \left(_\prod_{j=1}^{s} {C^{(m_{j}-|P|_{j})}_{j}(0)} \right) \, \gamma^{n-|P|+1} \prod_{W_\in P}{d_{W}}} {\displaystyle_\sum_{P \in \S} \kappa_{P} \pgf^{(|P|-1)} \prod_{W \in_P}{d_{W}}}.\end{aligned}$$ We thus have $$\begin{aligned} \frac{p_{k+1|k+1}(n)}{p_{k+1|k}(n)} &=_\frac{\displaystyle \sum_{\substack{P \in \S \\_|P| \leq_n+1}} _\left(\kappa_{P} \frac{n!}{(n-|P|+1)!} \gamma^{n-|P|+1}_\prod_{W \in_P}{d_{W}}\right)} _ {\displaystyle \sum

ely $g$ determines $\bigl(V,\tau\bigr)$ as an $N{=}1$ SVOA. This element $g$ lifts with order $2$ to ${\mathrm{Spin}}(24)$ and acts on the $24$-dimensional representation with spectrum $1^{12} (-1)^{12}$. ${\mathrm{Co}}_1$ has a unique conjugacy class of such elements, called “$2\mathrm{c}$” in GAP. This verifies the uniqueness of $\tau$ up to the $\operatorname{Aut}_{N{=}0}(V)$-action. The centralizer of $g$ in ${\mathrm{Co}}_1$ is $2^{11}{:}{\mathrm M}_{12}{:}2$. Quotienting by the central “$2$” to achieve a faithful action on $V$, we find that $\operatorname{Aut}_{N{=}1}(V) = 2^{10}{:}{\mathrm M}_{12}{:}2$, and that $\tau$ is the superconformal vector constructed in §\[exist.spin12.2\]. Uniqueness for ${\mathrm{Spin}}(12)_2$ {#unique.spin12} -------------------------------------- The (nonconformal) embedding ${\mathrm{Spin}}(12)_2 \subset {\mathrm{Spin}}(24)_1$ is unusual among WZW embeddings. Its level-rank dual is $\text{``}{\mathrm{Spin}}(2)_{12}\text{''} \subset {\mathrm U}(1)_{12}$, but that has as its coset all of ${\mathrm{SU}}(12)_1$, and not just its subalgebra ${\mathrm{Spin}}(12)_2$. The reason for this unusual behaviour is that the embedding ${\mathrm{Spin}}(12)_2 \subset {\mathrm{SU}}(12)_1$ is already conformal, and so has trivial coset. ${\mathrm{Spin}}(12)_2$ is the fixed points of an outer ${{\mathds}Z}_2$-action on ${\mathrm{SU}}(12)_1$. We showed in §\[subsec.inclusions\] that the VOA inclusion ${\mathrm{Spin}}(12)_2 \subset {\mathrm{SU}}(12)_1$ extends to an SVOA inclusion ${\mathrm{SO}}^+(12)_2 \subset 6{\mathrm{PSU}}(12)_1$. Let us repeat the argument: it suffices to study the spin-$\frac32$ fields,

ely $ g$ determines $ \bigl(V,\tau\bigr)$ as an $ N{=}1 $ SVOA. This element $ g$ lifts with order $ 2 $ to $ { \mathrm{Spin}}(24)$ and act on the $ 24$-dimensional theatrical performance with spectrum $ 1^{12 } (-1)^{12}$. $ { \mathrm{Co}}_1 $ has a unique conjugacy course of such elements, call “ $ 2\mathrm{c}$ ” in GAP. This verifies the uniqueness of $ \tau$ up to the $ \operatorname{Aut}_{N{=}0}(V)$-action. The centralizer of $ g$ in $ { \mathrm{Co}}_1 $ is $ 2^{11}{:}{\mathrm M}_{12}{:}2$. Quotienting by the central “ $ 2 $ ” to achieve a close natural process on $ V$, we find that $ \operatorname{Aut}_{N{=}1}(V) = 2^{10}{:}{\mathrm M}_{12}{:}2 $, and that $ \tau$ is the superconformal vector constructed in § \[exist.spin12.2\ ]. Uniqueness for $ { \mathrm{Spin}}(12)_2 $ { # unique.spin12 } -------------------------------------- The (nonconformal) embedding $ { \mathrm{Spin}}(12)_2 \subset { \mathrm{Spin}}(24)_1 $ is unusual among WZW embeddings. Its horizontal surface - rank dual is $ \text{``}{\mathrm{Spin}}(2)_{12}\text {" } \subset { \mathrm U}(1)_{12}$, but that has as its coset all of $ { \mathrm{SU}}(12)_1 $, and not precisely its subalgebra $ { \mathrm{Spin}}(12)_2$. The rationality for this unusual behaviour is that the embedding $ { \mathrm{Spin}}(12)_2 \subset { \mathrm{SU}}(12)_1 $ is already conformal, and so has fiddling coset. $ { \mathrm{Spin}}(12)_2 $ is the fixed points of an outer $ { { \mathds}Z}_2$-action on $ { \mathrm{SU}}(12)_1$. We show in § \[subsec.inclusions\ ] that the VOA inclusion $ { \mathrm{Spin}}(12)_2 \subset { \mathrm{SU}}(12)_1 $ extends to an SVOA inclusion $ { \mathrm{SO}}^+(12)_2 \subset 6{\mathrm{PSU}}(12)_1$. Let us duplicate the argument: it suffices to study the spin-$\frac32 $ fields,

ely $g$ determines $\bigl(V,\tau\binr)$ as an $N{=}1$ SVOA. Tkus elekent $g$ lifts wkth order $2$ to ${\mathrm{Spin}}(24)$ and axts ob the $24$-dimensional reprdsentatioj with spectcum $1^{12} (-1)^{12}$. ${\mathrm{Co}}_1$ hea a unique cohmugacv rlass of such ekements, canled “$2\mathrm{c}$” it EA'. This verifies the uniqueness of $\tat$ up to tje $\operatornamg{Aut}_{N{=}0}(N)$-astioh. The centralizer of $g$ in ${\mathrm{Co}}_1$ is $2^{11}{:}{\matirm M}_{12}{:}2$. Quotientimg by the central “$2$” to achifve w faithful action ln $V$, we fine thwr $\operatornaoe{Aut}_{N{=}1}(V) = 2^{10}{:}{\mathrm M}_{12}{:}2$, and jhat $\tau$ is the superconformal vdctor constructgb in §\[gxist.spin12.2\]. Uniwuenefs for ${\mathrm{Spin}}(12)_2$ {#unhque.spim12} -------------------------------------- The (nonconforkal) emvedding ${\mathrm{Spin}}(12)_2 \suuset {\mathrm{Spin}}(24)_1$ is ugusual ammny WZW embeddings. Its oecel-ratk dgal kw $\tdxt{``}{\javhrj{Spin}}(2)_{12}\tfxt{''} \subset {\mafhrm U}(1)_{12}$, but rhat has as its costt wol of ${\mathrm{SH}}(12)_1$, and goe just its subalgebra ${\mathrm{Spin}}(12)_2$. The redsoh for this unusual behaciour is that the embgdding ${\matrrm{Spin}}(12)_2 \subset {\mathrm{SU}}(12)_1$ is already conformal, and vo haa trnylal zisft. ${\mathrm{Spin}}(12)_2$ is the fixed points of an outer ${{\jauhdx}Z}_2$-action on ${\mabhrm{SU}}(12)_1$. We showed in §\[skbxgc.inclusions\] tfat thz VKA inclusion ${\mathrl{Spin}}(12)_2 \sobset {\nathrm{SU}}(12)_1$ txtencs to an SVOA inclusion ${\matyrm{SO}}^+(12)_2 \subset 6{\nathrm{PSU}}(12)_1$. Let us rzpeat the aryument: it siffices to study the spnn-$\frac32$ fields,

ely $g$ determines $\bigl(V,\tau\bigr)$ as an $N{=}1$ element lifts with $2$ to ${\mathrm{Spin}}(24)$ representation spectrum $1^{12} (-1)^{12}$. has a unique class of such elements, called “$2\mathrm{c}$” GAP. This verifies the uniqueness of $\tau$ up to the $\operatorname{Aut}_{N{=}0}(V)$-action. The centralizer $g$ in ${\mathrm{Co}}_1$ is $2^{11}{:}{\mathrm M}_{12}{:}2$. Quotienting by the central “$2$” to achieve faithful on we that $\operatorname{Aut}_{N{=}1}(V) = 2^{10}{:}{\mathrm M}_{12}{:}2$, and that $\tau$ is the superconformal vector constructed in §\[exist.spin12.2\]. Uniqueness ${\mathrm{Spin}}(12)_2$ {#unique.spin12} -------------------------------------- The (nonconformal) embedding ${\mathrm{Spin}}(12)_2 \subset is unusual among WZW Its level-rank dual is $\text{``}{\mathrm{Spin}}(2)_{12}\text{''} {\mathrm but that as coset of ${\mathrm{SU}}(12)_1$, and just its subalgebra ${\mathrm{Spin}}(12)_2$. The reason for this unusual behaviour is that the embedding ${\mathrm{Spin}}(12)_2 \subset {\mathrm{SU}}(12)_1$ already conformal, has trivial ${\mathrm{Spin}}(12)_2$ the points of an on ${\mathrm{SU}}(12)_1$. We showed in §\[subsec.inclusions\] inclusion ${\mathrm{Spin}}(12)_2 \subset {\mathrm{SU}}(12)_1$ extends to an SVOA ${\mathrm{SO}}^+(12)_2 \subset Let us repeat the argument: it to study the spin-$\frac32$ fields,

ely $g$ determines $\bigl(V,\tau\bigR)$ as an $N{=}1$ SVOA. this eLemEnt $G$ lIfts With Order $2$ to ${\mathrm{SPIn}}(24)$ anD acts on the $24$-dimensional rEpresEnTAtioN WiTh speCtrum $1^{12} (-1)^{12}$. ${\maTHrM{cO}}_1$ haS a UnIquE cONjUgacy ClaSs of sucH elements, cAllEd “$2\Mathrm{c}$” in GAP. tHiS verifies tHe uNiqueness of $\tAu$ uP to the $\OpEraTOrnamE{AuT}_{N{=}0}(V)$-acTion. ThE CentraLizer of $g$ iN ${\mAThrm{Co}}_1$ IS $2^{11}{:}{\mathrm m}_{12}{:}2$. qUoTienTing by the central “$2$” tO AcHIeve a faithful aCtion oN $V$, WE fIND thAt $\oPeratornamE{AUt}_{N{=}1}(V) = 2^{10}{:}{\mAThrm M}_{12}{:}2$, anD ThAT $\TAu$ iS The superconfoRmal vector cONstRucted In §\[ExiST.spin12.2\]. UNiqueNeSS foR ${\mathrm{Spin}}(12)_2$ {#UniqUe.spin12} -------------------------------------- The (NonconFOrmal) emBEdding ${\mAthrm{SPin}}(12)_2 \SubSet {\mAThRm{spiN}}(24)_1$ iS UnuSUaL amONg WzW embeddInGs. its leVel-rANK DUal iS $\teXt{``}{\maThrm{SPin}}(2)_{12}\text{''} \subset {\MatHrm U}(1)_{12}$, BUt tHat haS as itS cosEt All of ${\Mathrm{sU}}(12)_1$, and NoT just its subalgeBra ${\mAthrm{Spin}}(12)_2$. the ReAsoN fOr thiS UnusuaL beHavIour is tHat the eMBedDiNG ${\MAtHrm{Spin}}(12)_2 \subset {\mathrM{Su}}(12)_1$ IS aLready coNformaL, AnD sO Has triviAl CosEt. ${\maTHRm{SpiN}}(12)_2$ is tHE fIxed poinTs of an OUtEr ${{\Mathds}Z}_2$-AcTion on ${\MaThrM{SU}}(12)_1$. we shoWEd in §\[Subsec.InclusioNs\] thaT The VOA inclusioN ${\Mathrm{Spin}}(12)_2 \subSEt {\MAThRM{SU}}(12)_1$ eXteNds to an SVOA InclUSion ${\MathRM{So}}^+(12)_2 \suBSet 6{\maThrm{PsU}}(12)_1$. lEt US repeat the argument: iT sUfficeS to stUdy the spin-$\fraC32$ fields,

ely $g$ determines $\bigl( V,\tau\big r)$ a s a n $ N{ =}1$ SVO A. This eleme n t $g $ lifts with order $2$ to $ {\ m athr m {S pin}} (24)$ a n da c tson t he$2 4 $- dimen sio nal rep resentatio n w it h spectrum $ 1 ^{ 12} (-1)^{ 12} $. ${\mathrm {Co }}_1$ha s a uniqu e c onjug acy cl a ss ofsuch elem en t s, cal l ed “$2\ m a th rm{c }$” in GAP. Thisv er i fies the uniqu enessof $\ t a u$upto the $\o pe rator n ame{Aut } _{ N { = }0} ( V)$-action. T he centrali z erof $g$ i n $ { \mathr m{Co} }_ 1 $ i s $2^{11}{: }{\m athrm M}_ {12}{: } 2$. Quo t ienting by th e c ent ral“ $2 $” to a c hie v ea f a ith ful acti on o n $V$ , we f i n d th at$\op erato rname{Aut}_{N {=} 1}(V ) =2^{10 }{:}{ \mat hr m M}_ {12}{: }2$,an d that $\tau$ i s th e superco nfo rm alve ctorc onstru cte d i n §\[ex ist.spi n 12. 2\ ] . U niqueness for ${\m at h r m{ Spin}}(1 2)_2${ #u ni q ue.spin1 2} -- ---- - - ----- ---- - -- -------- ------ - -- - The (n on confor ma l)emb eddin g ${\ mathrm {Spin}}( 12)_2 \subset {\math r m{Spin}}(24)_ 1 $i s u n usua l a mong WZW em bedd i ngs. Its le vel - rankdualis $\ t ext{``}{\mathrm{Spi n} }(2)_{ 12}\t ext{''} \subs et {\mathr m U }(1)_{12 }$,b ut that has as it s cos et all of$ {\mathrm {SU}} (12)_1$, and notj u st its s uba lge bra ${ \ m at hrm{Spin}}(12 ) _ 2$.Th e reaso n f or this un usu albeh av iour is t hat theem be dd in g $ {\mat h rm{Spin} }( 12) _2 \s ubset {\math rm{SU }}(1 2) _1 $ is alread y c o n form al ,andsoha s tri vial cos et. ${\ mathrm{Sp in} } (12) _2 $is thefixed pointsof an outer${ {\m athds} Z } _2$-acti on on ${\mathrm{SU}}(12 ) _1$. W e s howed in§\[subsec .in clusio ns\ ] thatthe VO A inc lu sio n ${\ma t h rm {Sp in }}(12)_2 \ s u bse t {\m at hrm{ SU}}(12 )_1$ extends to an SVO A inclusion $ {\m athr m { SO }}^ + (1 2 )_2 \ s ubs e t 6{\mathrm{PSU} }(12)_1$.Le t u s repeat t h e a rg ument:it suff icest o study the spin -$\frac32 $fiel d s ,

ely $g$_determines $\bigl(V,\tau\bigr)$_as an $N{=}1$ SVOA. This_element $g$_lifts_with order_$2$_to ${\mathrm{Spin}}(24)$ and_acts on the_$24$-dimensional representation with spectrum_$1^{12} (-1)^{12}$. ${\mathrm{Co}}_1$_has_a unique conjugacy class of such elements, called “$2\mathrm{c}$” in GAP. This verifies the_uniqueness_of $\tau$_up_to_the $\operatorname{Aut}_{N{=}0}(V)$-action. The centralizer of_$g$ in ${\mathrm{Co}}_1$ is $2^{11}{:}{\mathrm_M}_{12}{:}2$. Quotienting_by the central “$2$” to achieve a faithful_action_on $V$, we_find that $\operatorname{Aut}_{N{=}1}(V) = 2^{10}{:}{\mathrm M}_{12}{:}2$, and that $\tau$_is the superconformal vector constructed in_§\[exist.spin12.2\]. Uniqueness for ${\mathrm{Spin}}(12)_2$_{#unique.spin12} -------------------------------------- The_(nonconformal)_embedding ${\mathrm{Spin}}(12)_2 \subset {\mathrm{Spin}}(24)_1$_is unusual among WZW embeddings. Its_level-rank dual is $\text{``}{\mathrm{Spin}}(2)_{12}\text{''} \subset {\mathrm_U}(1)_{12}$, but that has as its coset_all of ${\mathrm{SU}}(12)_1$, and not just_its subalgebra ${\mathrm{Spin}}(12)_2$. The reason_for this_unusual behaviour is that the_embedding ${\mathrm{Spin}}(12)_2 \subset_{\mathrm{SU}}(12)_1$ is_already conformal, and_so has trivial coset. ${\mathrm{Spin}}(12)_2$ is_the fixed points_of an outer ${{\mathds}Z}_2$-action on ${\mathrm{SU}}(12)_1$. We_showed_in §\[subsec.inclusions\] that_the_VOA_inclusion ${\mathrm{Spin}}(12)_2_\subset {\mathrm{SU}}(12)_1$ extends_to_an SVOA_inclusion_${\mathrm{SO}}^+(12)_2 \subset 6{\mathrm{PSU}}(12)_1$. Let us repeat_the_argument: it suffices to study the spin-$\frac32$_fields,

, u ) \in \mathcal{C} \times [0,1] : (\gamma_x, t, s, u ) \in \overline{\Pi}\}.$$Let us observe that $\overline{\Pi}^\varepsilon$ is a Poisson process on $\mathcal{C} \times [0,1]$ with intensity measure $\overline{\phi}_{\nu_\varepsilon}$ where $\nu_\varepsilon$ denotes the $\varepsilon$-*discretized intensity measure* defined by the formula $$\nu_\varepsilon := \nu \circ D_\varepsilon^{-1}.$$ Furthermore, establishes a one-to-one correspondence between cylinders of With this in mind, we shall write $C_\varepsilon$ to denote the $\varepsilon$-cylinder in $\Pi^\varepsilon$ which corresponds to the cylinder $C \in \Pi$, i.e., if $C=(\gamma_x,t,s)$ then we shall set $C_\varepsilon = (\gamma_x^\varepsilon,t,s )$. \[convaga\] For each $t \in {{\mathbb R}}$ we have $\Pi^\varepsilon_t \overset{as}{\longrightarrow} \Pi_t$ as $\varepsilon \rightarrow 0^+$ with the vague topology. Straightforward consequence of the following lemma. \[lemaconvaga\] Let $\xi \in \mathcal{N}(S \times G)$ and for each $\varepsilon > 0$ consider the configuration $\xi^\varepsilon$ defined by the standard representation $$\xi^\varepsilon = \sum_{\gamma_x \in Q_\xi} m(\gamma_x)\delta_{\gamma_x^\varepsilon}.\footnote{The fact that for every $\varepsilon > 0$ the configuration $\xi^\varepsilon$ is indeed \textit{locally finite} follows from the absolute local compactness of $S \times G$.}$$ Then with respect to the vague topology in $\mathcal{N}(S \times G)$ we have $\lim_{\varepsilon \rightarrow 0^+} \xi^\varepsilon = \xi.$ It suffices to show that for each compact set $K \subseteq S \times G$ and $\delta > 0$ there exists $\varepsilon_0 > 0$ small enough such that $\xi^\varepsilon \in (\xi)_{K,\delta}$ for all $0 <\varepsilon < \varepsilon_0$. Notice that if we take $$\varepsilon_0 := \left(\frac{1}{2}\min\{ d_{S \times G}\

, u) \in \mathcal{C } \times [ 0,1 ]: (\gamma_x, t, s, u) \in \overline{\Pi}\}.$$Let us observe that $ \overline{\Pi}^\varepsilon$ is a Poisson process on $ \mathcal{C } \times [ 0,1]$ with volume standard $ \overline{\phi}_{\nu_\varepsilon}$ where $ \nu_\varepsilon$ denotes the $ \varepsilon$-*discretized intensity measure * define by the formula $ $ \nu_\varepsilon: = \nu \circ D_\varepsilon^{-1}.$$ Furthermore, establishes a one - to - one parallelism between cylinder of With this in mind, we shall publish $ C_\varepsilon$ to denote the $ \varepsilon$-cylinder in $ \Pi^\varepsilon$ which corresponds to the cylinder $ vitamin c \in \Pi$, i.e., if $ C=(\gamma_x, t, s)$ then we shall set $ C_\varepsilon = (\gamma_x^\varepsilon, t, s) $. \[convaga\ ] For each $ t \in { { \mathbb R}}$ we have $ \Pi^\varepsilon_t \overset{as}{\longrightarrow } \Pi_t$ as $ \varepsilon \rightarrow 0^+$ with the vague topology. Straightforward consequence of the following lemma. \[lemaconvaga\ ] permit $ \xi \in \mathcal{N}(S \times G)$ and for each $ \varepsilon > 0 $ consider the configuration $ \xi^\varepsilon$ defined by the standard theatrical performance $ $ \xi^\varepsilon = \sum_{\gamma_x \in Q_\xi } m(\gamma_x)\delta_{\gamma_x^\varepsilon}.\footnote{The fact that for every $ \varepsilon > 0 $ the configuration $ \xi^\varepsilon$ is indeed \textit{locally finite } follows from the absolute local concentration of $ S \times G$.}$$ Then with respect to the undefined topology in $ \mathcal{N}(S \times G)$ we consume $ \lim_{\varepsilon \rightarrow 0^+ } \xi^\varepsilon = \xi.$ It suffices to show that for each compact dress $ K \subseteq S \times G$ and $ \delta > 0 $ there exists $ \varepsilon_0 > 0 $ small enough such that $ \xi^\varepsilon \in (\xi)_{K,\delta}$ for all $ 0 < \varepsilon < \varepsilon_0$. Notice that if we take $ $ \varepsilon_0: = \left(\frac{1}{2}\min\ { d_{S \times G}\

, u ) \in \mathcal{C} \times [0,1] : (\gamoa_x, t, s, u ) \in \orwrline{\'i}\}.$$Let ua observd that $\overline{\Pi}^\varepsilon$ ms a Poiswon process on $\mathcal{Z} \times [0,1]$ aith intwnsiuy measure $\overliis{\phi}_{\nu_\varepsimln}$ wkece $\nu_\varepsilon$ denotes tve $\varepsilon$-*dhszrztized intensity measure* defined by ehe forkupa $$\nu_\varepsilog := \nl \sirc D_\varepsilon^{-1}.$$ Furthermore, establishss a ont-to-one correspondemce between cylinders of Wlth hhis in mind, we shwll write $C_\care[wilon$ to denute the $\vagzpsilon$-cylihder in $\Pi^\varepsilon$ which corrdsponbs to the ciljnfgr $C \in \Pi$, i.x., if $C=(\damma_x,t,s)$ thek we shdll set $C_\varepsilon = (\namma_e^\varwpsilon,t,s )$. \[convaga\] For each $t \in {{\mathbb T}}$ we have $\[i^\rarepsilon_t \overset{as}{\oobgrigvtarsow} \Ou_t$ xs $\bacepailon \gigitarrow 0^+$ wifh the vaguw topology. Straightfprrqrd consequende of ehq following lemma. \[lemaconvaga\] Let $\xi \in \katgcal{N}(S \times G)$ and for wach $\varepsilon > 0$ condider the configuration $\xi^\varepsilon$ defined by the standasd re'rdseubwguoj $$\xi^\varepsilon = \sum_{\gamma_x \in Q_\xi} m(\gamma_x)\delta_{\dzmka_q^\varepsilon}.\footnobe{The fact that fot fvrty $\varepsilon > 0$ the eknriguration $\xi^\varepdilon$ if indwed \textiu{locakly finite} follows from the absolute lobal xompactness of $S \tnmes G$.}$$ Then cith rgspect to the vague topology nn $\matgcal{N}(S \timed G)$ we habd $\lim_{\varepsilon \figmtasrow 0^+} \xi^\varepsilon = \xi.$ It stffices ti shpw that for each sompact seh $K \smtseteq S \times G$ ajd $\deptd > 0$ there fxists $\varepsilon_0 > 0$ small enougi such that $\xo^\vdreisilon \in (\xi)_{K,\dclta}$ for all $0 <\vwrepsilon < \vargpsilon_0$. Njtice that if wv take $$\vacepsilon_0 := \lest(\frac{1}{2}\min\{ d_{S \dlmes G}\

, u ) \in \mathcal{C} \times [0,1] t, u ) \overline{\Pi}\}.$$Let us observe process $\mathcal{C} \times [0,1]$ intensity measure $\overline{\phi}_{\nu_\varepsilon}$ $\nu_\varepsilon$ denotes the $\varepsilon$-*discretized intensity measure* by the formula $$\nu_\varepsilon := \nu \circ D_\varepsilon^{-1}.$$ Furthermore, establishes a one-to-one correspondence cylinders of With this in mind, we shall write $C_\varepsilon$ to denote the in which to cylinder $C \in \Pi$, i.e., if $C=(\gamma_x,t,s)$ then we shall set $C_\varepsilon = (\gamma_x^\varepsilon,t,s )$. \[convaga\] each $t \in {{\mathbb R}}$ we have $\Pi^\varepsilon_t \Pi_t$ as $\varepsilon \rightarrow with the vague topology. Straightforward of following lemma. Let \in \times G)$ and each $\varepsilon > 0$ consider the configuration $\xi^\varepsilon$ defined by the standard representation $$\xi^\varepsilon = \sum_{\gamma_x \in m(\gamma_x)\delta_{\gamma_x^\varepsilon}.\footnote{The fact every $\varepsilon 0$ configuration is indeed \textit{locally from the absolute local compactness of Then with respect to the vague topology in \times G)$ have $\lim_{\varepsilon \rightarrow 0^+} \xi^\varepsilon = It suffices to show that for each compact $K \subseteq S \times G$ and $\delta > 0$ there exists $\varepsilon_0 > 0$ small that $\xi^\varepsilon \in (\xi)_{K,\delta}$ all $0 <\varepsilon \varepsilon_0$. that we $$\varepsilon_0 := d_{S \times G}\

, u ) \in \mathcal{C} \times [0,1] : (\gamma_x, t, s, u ) \In \overline{\pi}\}.$$Let Us oBseRvE thaT $\oveRline{\Pi}^\varepsiLOn$ is A Poisson process on $\mathcAl{C} \tiMeS [0,1]$ With INtEnsitY measurE $\OvERLinE{\pHi}_{\Nu_\vArEPsIlon}$ wHerE $\nu_\varePsilon$ denoTes ThE $\varepsilon$-*dIScRetized intEnsIty measure* deFinEd by thE fOrmULa $$\nu_\vArePsiloN := \nu \cirC d_\varepSilon^{-1}.$$ FurtHeRMore, esTAblisheS A OnE-to-oNe correspondence bETwEEn cylinders of WIth thiS iN MiND, We sHalL write $C_\varEpSilon$ TO denote THe $\VAREpsILon$-cylinder in $\pi^\varepsiloN$ WhiCh corrEsPonDS to the CylinDeR $c \in \pi$, i.e., if $C=(\gammA_x,t,s)$ Then we shaLl set $C_\VArepsilON = (\gamma_x^\VarepsIloN,t,s )$. \[ConvAGa\] foR eaCh $T \In {{\mAThBb R}}$ WE haVe $\Pi^\varePsIlOn_t \ovErseT{AS}{\LOngrIghTarrOw} \Pi_t$ As $\varepsilon \rIghTarrOW 0^+$ wiTh the Vague TopoLoGy. StrAightfOrwarD cOnsequence of the FollOwing lemmA. \[leMaConVaGa\] Let $\XI \in \matHcaL{N}(S \Times G)$ aNd for eaCH $\vaRePSILoN > 0$ consider the configUrATIoN $\xi^\varepSilon$ dEFiNeD By the staNdArd ReprESEntatIon $$\xI^\VaRepsilon = \Sum_{\gamMA_x \In q_\xi} m(\gamMa_X)\delta_{\GaMma_X^\vaRepsiLOn}.\foOtnote{the fact tHat foR Every $\varepsiloN > 0$ The configuratIOn $\XI^\VaREpsiLon$ Is indeed \texTit{lOCallY finITe} FolLOws frOm the AbSOlUTe local compactness oF $S \Times G$.}$$ then wIth respect to tHe vague topOLOGy in $\mathCal{N}(s \TiMEs G)$ we have $\lim_{\vaRepsiLon \rightarROw 0^+} \xi^\varePsiloN = \xi.$ It sufFices to shOW That for eAch ComPacT seT $k \SuBseteq S \times G$ AND $\delTa > 0$ There exIstS $\varepsIloN_0 > 0$ smAll EnoUgH such that $\Xi^\varepsIlOn \In (\Xi)_{k,\deLta}$ foR All $0 <\varepSiLon < \VaRepSilon_0$. nOtice tHat if We taKe $$\VaREpsIlon_0 := \lefT(\FrAC{1}{2}\Min\{ d_{s \tImEs G}\

, u ) \in \mathcal{C} \tim es [0,1] : (\ga mma _x, t , s, u ) \in \overline { \Pi} \}.$$Let us observe th at $\ ov e rlin e {\ Pi}^\ varepsi l on $ isaPo iss on pr ocess on $\math cal{C} \ti mes [ 0,1]$ with i n te nsity meas ure $\overline{ \ph i}_{\n u_ \va r epsil on} $ whe re $\n u _\vare psilon$ d en o tes th e $\vare p s il on$- *discretized inte n si t y measure* def ined b yt he f orm ula $$\nu_\va re psilo n := \nu \c i r c D_ \ varepsilon^{- 1}.$$ Furth e rmo re, es ta bli s hes aone-t o- o necorresponde ncebetween c ylinde r s of Wi t h thisin min d,weshal l w ri te$C _ \va r ep sil o n$to denot eth e $\v arep s i l o n$-c yli nder in $ \Pi^\varepsil on$ whi c h c orres ponds toth e cyl inder$C \i n\Pi$, i.e., if$C=( \gamma_x, t,s )$ th en we s h all se t $ C_\ varepsi lon = ( \ gam ma _ x ^ \v arepsilon,t,s )$. \ [ c on vaga\] F or eac h $ t\ in {{\ma th bbR}}$ w e hav e $\ P i^ \varepsi lon_t\ ov er set{as} {\ longri gh tar row } \Pi _ t$ a s $\va repsilon \rig h tarrow 0^+$ wi t h the vague t o po l o gy . St rai ghtforwardcons e quen ce o f t hef ollow ing l em m a. \[lemaconvaga\] Le t$\xi \ in \m athcal{N}(S \ times G)$a n d for eac h $\ v ar e psilon > 0$ co nside r the conf i guration $\xi ^\vareps ilon$ def i n ed by th e s tan dar d r e p re sentation $$\ x i ^\va re psilon= \ sum_{\g amm a_x \i n Q _\ xi} m(\ga mma_x)\d el ta _{ \g amm a_x^\ v arepsilo n} .\f oo tno te{Th e factthatforev er y $\ varepsi l on > 0$th econf igu ra tion$\xi ^ \va repsilo n$ is ind eed \tex ti t{ locally finite} foll ow s from the a bso lute l o c al compa ctness of $S \times G$. } $$ Then wi th re spec t to thevag ue top olo g y in $ \mathc al{N} (S \t i m es G) $ we ha ve $\lim_{\v a r eps ilon\r ight arrow 0 ^+} \xi^\varepsilo n =\xi.$ It suf fic es t o sh owt ha t fo re ach c ompact set $K \ subseteq S \ t im es G$ and$ \de lt a > 0$there e xists $\varep silon_0 > 0$ small e noug h suc h that $\x i^\varep silon \in (\xi) _ {K ,\del ta} $ foral l $ 0 <\v arepsi l on< \va repsil on _0$. N otice t hat if w e take $$\varepsilon_0:= \le ft(\f rac {1}{2}\mi n\{ d_{ S \timesG}\

, u_) \in_\mathcal{C} \times [0,1] :_(\gamma_x, t,_s,_u )_\in_\overline{\Pi}\}.$$Let us observe_that $\overline{\Pi}^\varepsilon$ is_a Poisson process on_$\mathcal{C} \times [0,1]$_with_intensity measure $\overline{\phi}_{\nu_\varepsilon}$ where $\nu_\varepsilon$ denotes the $\varepsilon$-*discretized intensity measure* defined by the formula_$$\nu_\varepsilon_:= \nu_\circ_D_\varepsilon^{-1}.$$_Furthermore, establishes a one-to-one correspondence_between cylinders of With this_in mind,_we shall write $C_\varepsilon$ to denote the $\varepsilon$-cylinder_in_$\Pi^\varepsilon$ which corresponds_to the cylinder $C \in \Pi$, i.e., if $C=(\gamma_x,t,s)$_then we shall set $C_\varepsilon =_(\gamma_x^\varepsilon,t,s )$. \[convaga\] For_each_$t_\in {{\mathbb R}}$ we_have $\Pi^\varepsilon_t \overset{as}{\longrightarrow} \Pi_t$ as $\varepsilon_\rightarrow 0^+$ with the vague topology. Straightforward_consequence of the following lemma. \[lemaconvaga\] Let $\xi_\in \mathcal{N}(S \times G)$ and for_each $\varepsilon > 0$ consider_the configuration_$\xi^\varepsilon$ defined by the standard_representation $$\xi^\varepsilon =_\sum_{\gamma_x \in_Q_\xi} m(\gamma_x)\delta_{\gamma_x^\varepsilon}.\footnote{The fact_that for every $\varepsilon > 0$_the configuration $\xi^\varepsilon$_is indeed \textit{locally finite} follows from_the_absolute local compactness_of_$S_\times G$.}$$_Then with respect_to_the vague_topology_in $\mathcal{N}(S \times G)$ we have_$\lim_{\varepsilon_\rightarrow 0^+} \xi^\varepsilon = \xi.$ It suffices to_show that for each_compact_set $K \subseteq S_\times G$ and $\delta >_0$ there exists $\varepsilon_0 > 0$_small enough_such that_$\xi^\varepsilon \in (\xi)_{K,\delta}$ for all $0 <\varepsilon < \varepsilon_0$. Notice that_if we take $$\varepsilon_0 := \left(\frac{1}{2}\min\{_d_{S \times G}\

lim_{t\to\infty}\frac{R_7(t)\int_{0}^{t}g_2(s)dB(s)}{\sqrt{2t\log\log t}} =0, \quad \text{a.s.}$$ Similarly it may be shown that $$\label{eq:R8g4} \lim_{t\to\infty}\frac{R_8(t)\int_{0}^{t}g_4(s)dB(s)}{\sqrt{2t\log\log t}} =0, \quad \text{a.s.}$$ Then due to,,,, and Lemma \[lm:uplil\], by taking the limit superior across we get $$\limsup_{t\to\infty}\frac{X(t)}{\sqrt{2t\log\log t}} = \frac{\sigma}{\sqrt{3}}, \quad \text{a.s.}$$ Taking the limit inferior and applying these preparatory estimates along with Lemma \[lm:uplil\] secures the corresponding limit inferior. By and we have that $$\begin{gathered} \label{eq.Varbessel} \frac{1}{\sigma^2}\text{Var}[X(t)]=\int_0^t r(t,s)^2\,ds \\ = r_7(t)^2\int_0^t d_7(s)^2\,ds + 2r_7(t)r_8(t)\int_0^t d_7(s)d_8(s)\,ds + r_8(t)^2\int_0^t d_8(s)^2\,ds.\end{gathered}$$ We deduce the asymptotic behaviour of the terms on the righthand side of. By the definition of $d_7$ we have the identity $$\frac{1}{t^{3/2}}\int_0^t d_7(s)^2\,ds - \frac{1}{t^{3/2}}\int_0^t g_2^2(s)\,ds = 2\frac{1}{t^{3/2}}\int_0^t g_2(s)D_7(s)\,ds +\frac{1}{t^{3/2}}\int_0^t D_7(s)^2.$$ By the definition of $g_2

lim_{t\to\infty}\frac{R_7(t)\int_{0}^{t}g_2(s)dB(s)}{\sqrt{2t\log\log t } } = 0, \quad \text{a.s.}$$ Similarly it may be shown that $ $ \label{eq: R8g4 } \lim_{t\to\infty}\frac{R_8(t)\int_{0}^{t}g_4(s)dB(s)}{\sqrt{2t\log\log t } } = 0, \quad \text{a.s.}$$ Then due to, ,, , and Lemma   \[lm: uplil\ ], by taking the limit superior across we draw $ $ \limsup_{t\to\infty}\frac{X(t)}{\sqrt{2t\log\log metric ton } } = \frac{\sigma}{\sqrt{3 } }, \quad \text{a.s.}$$ Taking the limit inferior and apply these preparatory estimates along with Lemma   \[lm: uplil\ ] secures the represent terminus ad quem inferior. By and we have that $ $ \begin{gathered } \label{eq. Varbessel } \frac{1}{\sigma^2}\text{Var}[X(t)]=\int_0^t r(t, s)^2\,ds \\ = r_7(t)^2\int_0^t d_7(s)^2\,ds + 2r_7(t)r_8(t)\int_0^t d_7(s)d_8(s)\,ds + r_8(t)^2\int_0^t d_8(s)^2\,ds.\end{gathered}$$ We deduce the asymptotic behaviour of the terms on the righthand english of. By the definition of $ d_7 $ we have the identity $ $ \frac{1}{t^{3/2}}\int_0^t d_7(s)^2\,ds - \frac{1}{t^{3/2}}\int_0^t g_2 ^ 2(s)\,ds = 2\frac{1}{t^{3/2}}\int_0^t g_2(s)D_7(s)\,ds + \frac{1}{t^{3/2}}\int_0^t D_7(s)^2.$$ By the definition of $ g_2

lim_{h\to\infty}\frac{R_7(t)\int_{0}^{t}g_2(s)dB(s)}{\sdrt{2t\log\log t}} =0, \qoae \text{e.s.}$$ Simimarly it may be shown that $$\label{eq:R8g4} \lim_{u\no\infty}\frac{R_8(t)\int_{0}^{t}g_4(s)dB(r)}{\sqrt{2t\log\pog t}} =0, \qyad \uext{a.s.}$$ Then due to,,,, and Lemma \[lm:ulpil\], yy taking the likit superimr across we gat $$\lnmsup_{t\to\infty}\frac{X(t)}{\sqrt{2t\log\log t}} = \frwc{\sigma}{\xqgt{3}}, \quad \text{a.s.}$$ Takpnd ths limit inferior and applying thess prepagatory estimates slong with Lemma \[lm:uplil\] sefured the correspondinh limit infgdiow. Vy and we haxe that $$\begin{gathered} \mabel{eq.Varbessel} \frac{1}{\sigma^2}\text{Vaf}[X(t)]=\inc_0^t r(t,s)^2\,ds \\ = r_7(j)^2\nbt_0^t g_7(s)^2\,ds + 2r_7(t)r_8(t)\iit_0^t d_7(s)q_8(s)\,ds + r_8(t)^2\int_0^t d_8(s)^2\,ds.\end{catherec}$$ We deduce thc asykptitic behaviour of the terms on the righjhand side oy. By the definition od $e_7$ we vave the udevtiuy $$\hrad{1}{t^{3/2}}\int_0^t d_7(s)^2\,vs - \frac{1}{t^{3/2}}\inf_0^t g_2^2(s)\,ds = 2\drac{1}{t^{3/2}}\int_0^t g_2(s)D_7(s)\,ds +\frsc{1}{e^{3/2}}\pmt_0^t D_7(s)^2.$$ By ths defigieion of $g_2

lim_{t\to\infty}\frac{R_7(t)\int_{0}^{t}g_2(s)dB(s)}{\sqrt{2t\log\log t}} =0, \quad \text{a.s.}$$ Similarly it shown $$\label{eq:R8g4} \lim_{t\to\infty}\frac{R_8(t)\int_{0}^{t}g_4(s)dB(s)}{\sqrt{2t\log\log =0, \quad \text{a.s.}$$ \[lm:uplil\], taking the limit across we get t}} = \frac{\sigma}{\sqrt{3}}, \quad \text{a.s.}$$ Taking limit inferior and applying these preparatory estimates along with Lemma \[lm:uplil\] secures the limit inferior. By and we have that $$\begin{gathered} \label{eq.Varbessel} \frac{1}{\sigma^2}\text{Var}[X(t)]=\int_0^t r(t,s)^2\,ds \\ = d_7(s)^2\,ds 2r_7(t)r_8(t)\int_0^t + d_8(s)^2\,ds.\end{gathered}$$ We deduce the asymptotic behaviour of the terms on the righthand side of. By the of $d_7$ we have the identity $$\frac{1}{t^{3/2}}\int_0^t d_7(s)^2\,ds \frac{1}{t^{3/2}}\int_0^t g_2^2(s)\,ds = 2\frac{1}{t^{3/2}}\int_0^t +\frac{1}{t^{3/2}}\int_0^t D_7(s)^2.$$ By the definition $g_2

lim_{t\to\infty}\frac{R_7(t)\int_{0}^{t}g_2(s)dB(S)}{\sqrt{2t\log\lOg t}} =0, \quAd \tExt{A.s.}$$ simiLarlY it may be shown tHAt $$\laBel{eq:R8g4} \lim_{t\to\infty}\frac{r_8(t)\int_{0}^{T}g_4(S)DB(s)}{\sQRt{2T\log\lOg t}} =0, \quad \TExT{A.S.}$$ ThEn DuE to,,,, AnD leMma \[lm:UplIl\], by takIng the limiT suPeRior across we GEt $$\Limsup_{t\to\iNftY}\frac{X(t)}{\sqrt{2t\Log\Log t}} = \frAc{\SigMA}{\sqrt{3}}, \QuaD \text{A.s.}$$ TakiNG the liMit inferiOr ANd applYIng thesE PReParaTory estimates alonG WiTH Lemma \[lm:uplil\] sEcures ThE CoRREspOndIng limit inFeRior. BY And we haVE tHAT $$\BegIN{gathered} \labeL{eq.VarbesseL} \FraC{1}{\sigma^2}\TeXt{VAR}[X(t)]=\int_0^T r(t,s)^2\,dS \\ = r_7(T)^2\Int_0^T d_7(s)^2\,ds + 2r_7(t)r_8(t)\inT_0^t d_7(s)D_8(s)\,ds + r_8(t)^2\int_0^T d_8(s)^2\,ds.\eND{gatherED}$$ We deduCe the aSymPtoTic bEHaViOur Of THe tERmS on THe rIghthand SiDe Of. By tHe deFINITion Of $d_7$ We haVe the Identity $$\frac{1}{t^{3/2}}\Int_0^T d_7(s)^2\,dS - \FraC{1}{t^{3/2}}\int_0^T g_2^2(s)\,ds = 2\Frac{1}{T^{3/2}}\iNt_0^t g_2(s)d_7(s)\,ds +\frAc{1}{t^{3/2}}\inT_0^t d_7(s)^2.$$ By the definitiOn of $G_2

lim_{t\to\infty}\frac{R_7( t)\int_{0} ^{t}g _2( s)d B( s)}{ \sqr t{2t\log\log t } } =0 , \quad \text{a.s.}$$Simil ar l y it ma y beshown t h at $ $\l ab el {eq :R 8 g4 } \l im_{t\t o\infty}\f rac {R _8(t)\int_{0 } ^{ t}g_4(s)dB (s) }{\sqrt{2t\l og\ log t} }=0, \quad \t ext{a .s.}$$ Then d ue to,,,, a n d Lemm a  \[lm:u p l il \],by taking the lim i ts uperior across we ge t$ $\ l i msu p_{ t\to\infty }\ frac{ X (t)}{\s q rt { 2 t \lo g \log t}} = \f rac{\sigma} { \sq rt{3}} ,\qu a d \tex t{a.s .} $ $ T aking the l imit inferior and a p plyingt hese pr eparat ory es tima t es a lon gw ith Le mma \[l m:uplil\ ]se cures the c o r resp ond inglimit inferior. B y a nd w e ha ve th at $$ \beg in {gath ered} \lab el {eq.Varbessel}\fra c{1}{\sig ma^ 2} \te xt {Var} [ X(t)]= \in t_0 ^t r(t, s)^2\,d s \\ = r _ 7( t)^2\int_0^t d_7(s )^ 2 \ ,d s + 2r_7 (t)r_8 ( t) \i n t_0^t d_ 7( s)d _8(s ) \ ,ds + r_8 ( t) ^2\int_0 ^t d_8 ( s) ^2 \,ds.\e nd {gathe re d}$ $ W e ded u ce t he asy mptoticbehav i our of the ter m s on the righ t ha n d s i de o f.By the defi niti o n of $d_ 7 $weh ave t he id en t it y $$\frac{1}{t^{3/2} }\ int_0^ t d_7 (s)^2\,ds - \ frac{1}{t^ { 3 / 2}}\int_ 0^tg _2 ^ 2(s)\,ds = 2\fra c{1}{t^{3/ 2 }}\int_0 ^t g_ 2(s)D_7( s)\,ds +\ f r ac{1}{t^ {3/ 2}} \in t_0 ^ t D _7(s)^2.$$ By t he d ef inition of $g_2

lim_{t\to\infty}\frac{R_7(t)\int_{0}^{t}g_2(s)dB(s)}{\sqrt{2t\log\log t}}_=0, \quad_\text{a.s.}$$ Similarly it may_be shown_that_$$\label{eq:R8g4} __ \lim_{t\to\infty}\frac{R_8(t)\int_{0}^{t}g_4(s)dB(s)}{\sqrt{2t\log\log t}}_=0, \quad \text{a.s.}$$_Then due to,,,, and_Lemma \[lm:uplil\], by taking_the_limit superior across we get $$\limsup_{t\to\infty}\frac{X(t)}{\sqrt{2t\log\log t}} = \frac{\sigma}{\sqrt{3}}, \quad \text{a.s.}$$ Taking the limit_inferior_and applying_these_preparatory_estimates along with Lemma \[lm:uplil\] secures_the corresponding limit inferior. By and_we have_that $$\begin{gathered} \label{eq.Varbessel} \frac{1}{\sigma^2}\text{Var}[X(t)]=\int_0^t r(t,s)^2\,ds \\ = r_7(t)^2\int_0^t d_7(s)^2\,ds +_2r_7(t)r_8(t)\int_0^t_d_7(s)d_8(s)\,ds + r_8(t)^2\int_0^t_d_8(s)^2\,ds.\end{gathered}$$ We deduce the asymptotic behaviour of the terms_on the righthand side of. By_the definition of_$d_7$_we_have the identity $$\frac{1}{t^{3/2}}\int_0^t_d_7(s)^2\,ds - \frac{1}{t^{3/2}}\int_0^t g_2^2(s)\,ds =_ 2\frac{1}{t^{3/2}}\int_0^t g_2(s)D_7(s)\,ds +\frac{1}{t^{3/2}}\int_0^t D_7(s)^2.$$ By_the definition of $g_2

By Remark \[R1.2\], since $\operatorname{dim}\{ E(-1,\nabla)\}=3$, the corresponding Riemannian extension $\mathcal{N}:=(T^*M,g_{\nabla,\Phi})$ is conformally Einstein for any $\Phi$. Although in this instance $\mathcal{N}$ is locally conformally flat if $\Phi=0$ [@Afifi], $\mathcal{N}$ is not locally conformally flat for generic $\Phi\neq 0$. Type $\mathcal{B}$ surfaces {#ss1.6} --------------------------- Let $\mathfrak{B}$ be the unital non-commutative algebra generated by the vector fields $\partial_{x^2}$ and $X:=x^1\partial_{x^1}+x^2\partial_{x^2}$. The representation theory of the algebra $\mathfrak{B}$ is crucial to our investigation: \[T1.12\] Let $\mathcal{E}$ be a finite dimensional $\mathfrak{B}$ submodule of $C^\infty(\mathbb{R}^+\times\mathbb{R})\otimes_{\mathbb{R}}\mathbb{C}$. 1. If $f\in\mathcal{E}$, then $ f=\sum_{\alpha,i,j}c_{\alpha,i,j}(x^1)^\alpha(\log(x^1))^i(x^2)^j$ where in this finite sum $c_{\alpha,i,j}\in\mathbb{C}$, $\alpha\in\mathbb{C}$, and $i$ and $j$ are non-negative integers. 2. If $\dim\{\mathcal{E}\}=1$, then $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^a\}$ for some $a$. 3. If $\dim\{\mathcal{E}\}=2$, then one of the following possibilities hold: 1. $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha(c_1x^1+x^2)\}$. 2. $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha\log(x^1)\}$. 3. $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\beta\}$ for

By Remark   \[R1.2\ ], since $ \operatorname{dim}\ { E(-1,\nabla)\}=3 $, the corresponding Riemannian extension $ \mathcal{N}:=(T^*M, g_{\nabla,\Phi})$ is conformally Einstein for any $ \Phi$. Although in this instance $ \mathcal{N}$ is locally conformally flat if $ \Phi=0 $ [ @Afifi ], $ \mathcal{N}$ is not locally conformally bland for generic $ \Phi\neq 0$. character   $ \mathcal{B}$ surfaces { # ss1.6 } --------------------------- Let $ \mathfrak{B}$ be the unital non - commutative algebra generate by the vector battlefield $ \partial_{x^2}$ and $ X:=x^1\partial_{x^1}+x^2\partial_{x^2}$. The representation theory of the algebra $ \mathfrak{B}$ is all-important to our investigation: \[T1.12\ ] Let $ \mathcal{E}$ be a finite dimensional $ \mathfrak{B}$ submodule of $ C^\infty(\mathbb{R}^+\times\mathbb{R})\otimes_{\mathbb{R}}\mathbb{C}$. 1. If $ f\in\mathcal{E}$, then $ f=\sum_{\alpha, i, j}c_{\alpha, i, j}(x^1)^\alpha(\log(x^1))^i(x^2)^j$ where in this finite sum $ c_{\alpha, i, j}\in\mathbb{C}$, $ \alpha\in\mathbb{C}$, and $ i$ and $ j$ are non - damaging integers. 2. If $ \dim\{\mathcal{E}\}=1 $, then $ \mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^a\}$ for some $ a$. 3. If $ \dim\{\mathcal{E}\}=2 $, then one of the following possibility hold: 1. $ \mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha(c_1x^1+x^2)\}$. 2. $ \mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha\log(x^1)\}$. 3. $ \mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\beta\}$ for

By Remark \[R1.2\], since $\operatornxme{dim}\{ E(-1,\nabla)\}=3$, tkw corrxspondihg Riemavnian extension $\mathcal{N}:=(T^*M,g_{\nebla,\Phi})$ iw conformally Einstein for any $\Ihi$. Althoygh mn this instance $\mathcal{K}$ is mlcalnb conformally fkat if $\Phi=0$ [@Afifi], $\mathcal{T}$ ks not locally conformally flat for gqneric $\Lhl\neq 0$. Type $\mathcwl{B}$ xtrfadvs {#ss1.6} --------------------------- Let $\mathfrak{B}$ be the unitam non-cokmutative algrbra generated by the vectlr flelds $\partial_{x^2}$ and $X:=x^1\partial_{x^1}+z^2\pareual_{x^2}$. The repfesentation theory of jhe algebra $\mathfrak{B}$ is crucial to obr investigqtuon: \[H1.12\] Let $\mathcao{E}$ be a finite dimensionan $\mathftak{B}$ submodule of $C^\ibfty(\mathbb{R}^+\times\mathbu{R})\otimes_{\mathbb{R}}\mathbf{C}$. 1. If $f\it\mcthcal{E}$, then $ f=\sum_{\alpya,u,j}c_{\alkha,i,j}(f^1)^\alpfq(\loe(x^1))^i(s^2)^j$ wgere ij tiis finite aum $c_{\alpha,i,h}\in\mathbb{C}$, $\alpha\in\mstrvb{C}$, and $i$ and $j$ are njn-negative integers. 2. If $\dim\{\mathcal{E}\}=1$, thet $\mzthcal{E}=\operatorname{Span}_{\nathbb{C}}\{(x^1)^a\}$ for some $a$. 3. If $\dim\{\maehcal{E}\}=2$, then one of the following possibilities hond: 1. $\mcbmcal{D}=\ipfratorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha(c_1x^1+x^2)\}$. 2. $\matrdak{E}=\pperatorname{Spcn}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\slohs\jog(x^1)\}$. 3. $\matfcal{E}=\o'srztorname{Span}_{\mathbb{F}}\{(x^1)^\alpha,(v^1)^\beta\}$ for

By Remark \[R1.2\], since $\operatorname{dim}\{ E(-1,\nabla)\}=3$, the extension is conformally for any $\Phi$. is conformally flat if [@Afifi], $\mathcal{N}$ is locally conformally flat for generic $\Phi\neq Type $\mathcal{B}$ surfaces {#ss1.6} --------------------------- Let $\mathfrak{B}$ be the unital non-commutative algebra generated the vector fields $\partial_{x^2}$ and $X:=x^1\partial_{x^1}+x^2\partial_{x^2}$. The representation theory of the algebra $\mathfrak{B}$ crucial our \[T1.12\] $\mathcal{E}$ be a finite dimensional $\mathfrak{B}$ submodule of $C^\infty(\mathbb{R}^+\times\mathbb{R})\otimes_{\mathbb{R}}\mathbb{C}$. 1. If $f\in\mathcal{E}$, then $ f=\sum_{\alpha,i,j}c_{\alpha,i,j}(x^1)^\alpha(\log(x^1))^i(x^2)^j$ where this finite sum $c_{\alpha,i,j}\in\mathbb{C}$, $\alpha\in\mathbb{C}$, and $i$ and are non-negative integers. 2. $\dim\{\mathcal{E}\}=1$, then $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^a\}$ for some 3. $\dim\{\mathcal{E}\}=2$, then of following hold: 1. $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha(c_1x^1+x^2)\}$. $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha\log(x^1)\}$. 3. $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\beta\}$ for

By Remark \[R1.2\], since $\operatornamE{dim}\{ E(-1,\nabla)\}=3$, The coRreSpoNdIng RIemaNnian extension $\MAthcAl{N}:=(T^*M,g_{\nabla,\Phi})$ is conforMally eiNSteiN FoR any $\PHi$. AlthoUGh IN ThiS iNsTanCe $\MAtHcal{N}$ Is lOcally cOnformally FlaT iF $\Phi=0$ [@Afifi], $\matHCaL{N}$ is not locAllY conformally FlaT for geNeRic $\pHi\neq 0$. typE $\mathCal{B}$ suRFaces {#sS1.6} --------------------------- Let $\mathfRaK{b}$ be the UNital noN-COmMutaTive algebra generaTEd BY the vector fielDs $\partIaL_{X^2}$ aND $x:=x^1\pArtIal_{x^1}+x^2\partiAl_{X^2}$. The rEPresentATiON THeoRY of the algebra $\Mathfrak{B}$ is CRucIal to oUr InvEStigatIon: \[T1.12\] LEt $\MAthCal{E}$ be a finiTe diMensional $\MathfrAK{B}$ submoDUle of $C^\iNfty(\maThbB{R}^+\tImes\MAtHbB{R})\oTiMEs_{\mAThBb{R}}\MAthBb{C}$. 1. If $f\in\MaThCal{E}$, tHen $ f=\SUM_{\ALpha,I,j}c_{\AlphA,i,j}(x^1)^\aLpha(\log(x^1))^i(x^2)^j$ whEre In thIS fiNite sUm $c_{\alPha,i,J}\iN\mathBb{C}$, $\alpHa\in\mAtHbb{C}$, and $i$ and $j$ are Non-nEgative inTegErS. 2. If $\DiM\{\mathCAl{E}\}=1$, theN $\maThcAl{E}=\operAtornamE{spaN}_{\mATHBb{c}}\{(x^1)^a\}$ for some $a$. 3. If $\dim\{\maThCAL{E}\}=2$, Then one oF the foLLoWiNG possibiLiTieS holD: 1. $\MAthcaL{E}=\opERaTorname{SPan}_{\matHBb{c}}\{(x^1)^\Alpha,(x^1)^\aLpHa(c_1x^1+x^2)\}$. 2. $\mAtHcaL{E}=\oPeratORnamE{Span}_{\mAthbb{C}}\{(x^1)^\aLpha,(x^1)^\ALpha\log(x^1)\}$. 3. $\mathcaL{e}=\operatorname{sPaN}_{\MAtHBb{C}}\{(x^1)^\AlpHa,(x^1)^\beta\}$ for

By Remark \[R1.2\], since $\operato rname {di m}\ {E(-1 ,\na bla)\}=3$, the corr esponding Riemannian e xtens io n $\m a th cal{N }:=(T^* M ,g _ { \na bl a, \Ph i} ) $is co nfo rmallyEinstein f oran y $\Phi$. Al t ho ugh in thi s i nstance $\ma thc al{N}$ i s l o cally co nform ally f l at if$\Phi=0$[@ A fifi], $\mathc a l {N }$ i s not locally con f or m ally flat forgeneri c$ \P h i \ne q 0 $. Type $ \m athca l {B}$ su r fa c e s {# s s1.6} ------- ----------- - --- ----- L et$ \mathf rak{B }$ bethe unitalnon- commutati ve alg e bra gen e rated b y thevec tor fie l ds $ \pa rt i al_ { x^ 2}$ and $X:=x^1 \p ar tial_ {x^1 } + x ^ 2\pa rti al_{ x^2}$ . The represe nta tion the ory o f the alg eb ra $\ mathfr ak{B} $is crucial to o ur i nvestigat ion : \[ T1 .12\] Let $\ mat hca l{E}$ b e a fin i tedi m e n si onal $\mathfrak{B} $s u bm odule of $C^\i n ft y( \ mathbb{R }^ +\t imes \ m athbb {R}) \ ot imes_{\m athbb{ R }} \m athbb{C }$ . 1. I f $ f\i n\mat h cal{ E}$, t hen $ f= \sum_ { \alpha,i,j}c_{ \ alpha,i,j}(x^ 1 )^ \ a lp h a(\l og( x^1))^i(x^2 )^j$ wher e in th isf inite sum$c _ {\ a lpha,i,j}\in\mathbb {C }$, $\ alpha \in\mathbb{C} $, and $i$ a n d $j$ ar e no n -n e gative integer s. 2 . If $\di m \{\mathc al{E} \}=1$, t hen $\mat h c al{E}=\o per ato rna me{ S p an }_{\mathbb{C} } \ {(x^ 1) ^a\}$ f orsome $a $. 3. I f $ \d im\{\math cal{E}\} =2 $, t he n o ne of the foll ow ing p oss ibili t ies ho ld: 1. $ \ma thcal{E } =\ o p erat or na me{S pan }_ {\mat hbb{ C }}\ {(x^1)^ \alpha,(x ^1) ^ \alp ha (c _1x^1+x ^2)\}$. 2. $\mathca l{ E}= \opera t o rname{Sp an}_{\mathbb{C}}\{(x^1) ^ \alpha, (x^ 1)^\a lpha \log(x^1) \}$ . 3. $\mat hcal{E }=\op er ato r n ame{S p a n} _{\ ma thbb{C}}\{ ( x ^1) ^\alp ha ,(x^ 1)^\bet a\}$ for

By_Remark \[R1.2\], since_$\operatorname{dim}\{ E(-1,\nabla)\}=3$, the corresponding_Riemannian extension_$\mathcal{N}:=(T^*M,g_{\nabla,\Phi})$_is conformally_Einstein_for any $\Phi$._Although in this_instance $\mathcal{N}$ is locally_conformally flat if_$\Phi=0$_[@Afifi], $\mathcal{N}$ is not locally conformally flat for generic $\Phi\neq 0$. Type $\mathcal{B}$ surfaces {#ss1.6} --------------------------- Let $\mathfrak{B}$_be_the unital_non-commutative_algebra_generated by the vector fields_$\partial_{x^2}$ and $X:=x^1\partial_{x^1}+x^2\partial_{x^2}$. The representation_theory of_the algebra $\mathfrak{B}$ is crucial to our investigation: \[T1.12\]_Let_$\mathcal{E}$ be a_finite dimensional $\mathfrak{B}$ submodule of $C^\infty(\mathbb{R}^+\times\mathbb{R})\otimes_{\mathbb{R}}\mathbb{C}$. 1. If $f\in\mathcal{E}$,_then $ f=\sum_{\alpha,i,j}c_{\alpha,i,j}(x^1)^\alpha(\log(x^1))^i(x^2)^j$ where in this_finite sum $c_{\alpha,i,j}\in\mathbb{C}$,_$\alpha\in\mathbb{C}$,_and_$i$ and $j$ are_non-negative integers. 2. If $\dim\{\mathcal{E}\}=1$, then_$\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^a\}$ for some $a$. 3. If_$\dim\{\mathcal{E}\}=2$, then one of the following possibilities_hold: 1. _$\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha(c_1x^1+x^2)\}$. 2._ $\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\alpha\log(x^1)\}$. _ 3. _$\mathcal{E}=\operatorname{Span}_{\mathbb{C}}\{(x^1)^\alpha,(x^1)^\beta\}$ for

{\ensuremath{\{z \in \C \,\colon |z| = 1\}}}$, and let ${\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$ denote the class of all liftings of continuous circle maps of degree one. If $F\in {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, $\operatorname{Rot}(F)$ denotes the rotation set of $F$ and, by [@Ito], is a compact non empty interval. To study the connection between the set of periods and the rotation interval, we need some additional notation. For all $c\le d$, we set $M(c,d):={\ensuremath{\{n \in \N \,\colon c < k/n < d \text{ for some integer $k$}\}}}$. Notice that we do not assume here that $k$ and $n$ are coprime. Obviously, $M(c,d)= \emptyset$ if and only if $c=d$. Given $\rho \in \R$ and $S \subset \N$, we set $$\Lambda(\rho, S) = \begin{cases} \emptyset & \text{if $\rho \notin \Q$}, \\ {\ensuremath{\{nq \,\colon q\in S\}}} & \text{if $\rho=k/n$ with $k$ and $n$ coprime}. \end{cases}$$ The next theorem recalls Misiurewicz’s characterization of the sets of periods for degree $1$ circle maps (see [@Mis; @ALM]). \[S9.5\] Let $F\in {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, and let $\operatorname{Rot}(F) = [c,d]$. Then there exist numbers $s_c,s_d \in \N_{\Sho}$ such that $\operatorname{Per}(F)= \Lambda(c,\operatorname{S\mbox{\tiny\textup{sh}}}(s_c)) \cup M(c,d) \cup \Lambda(d,\operatorname{S\mbox{\tiny\textup{sh}}}(s_d))$. Conversely, for all $c,d \in \R$ with $c \le d$ and all $s_c, s_d\in\N_{\Sho}$, there exists a map $F \in {\mathcal{L}}_1

{ \ensuremath{\{z \in \C \,\colon |z| = 1\}}}$, and let $ { \mathcal{L}}_1({\ensuremath{\mathbb{R}}})$ denote the class of all liftings of continuous circle map of academic degree one. If $ F\in { \mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, $ \operatorname{Rot}(F)$ denotes the rotation set of $ F$ and, by [ @Ito ], is a compendious non empty interval. To learn the connection between the set of period and the rotation interval, we need some extra note. For all $ c\le d$, we set $ M(c, d):={\ensuremath{\{n \in \N \,\colon c < k / n < d \text { for some integer $ k$}\}}}$. Notice that we do not wear here that $ k$ and $ n$ are coprime. Obviously, $ M(c, d)= \emptyset$ if and only if $ c = d$. give $ \rho \in \R$ and $ S \subset \N$, we set $ $ \Lambda(\rho, S) = \begin{cases } \emptyset & \text{if $ \rho \notin \Q$ }, \\ { \ensuremath{\{nq \,\colon q\in S\ } } } & \text{if $ \rho = k / n$ with $ k$ and $ n$ coprime }. \end{cases}$$ The next theorem recalls Misiurewicz ’s word picture of the sets of periods for degree $ 1 $ r-2 maps (see [ @Mis; @ALM ]). \[S9.5\ ] Let $ F\in { \mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, and let $ \operatorname{Rot}(F) = [ degree centigrade, d]$. Then there exist numbers $ s_c, s_d \in \N_{\Sho}$ such that $ \operatorname{Per}(F)= \Lambda(c,\operatorname{S\mbox{\tiny\textup{sh}}}(s_c) ) \cup M(c, d) \cup \Lambda(d,\operatorname{S\mbox{\tiny\textup{sh}}}(s_d))$. Conversely, for all $ c, d \in \R$ with $ c \le d$ and all $ s_c, s_d\in\N_{\Sho}$, there exists a map $ F \in { \mathcal{L}}_1

{\enskremath{\{z \in \C \,\colon |z| = 1\}}}$, xnd let ${\mathcal{L}}_1({\ensurekath{\mafhbb{R}}})$ devote the class of all liftinjs od conupnuous circle maps of degree oje. If $F\ib {\matical{L}}_1({\ensuremath{\mefhbb{R}}})$, $\oizratodkame{Rmv}(F)$ denotes the totation set of $F$ and, by [@Ido], id a compact non empty interval. To sttdy the clnnection betwgen tnq sef of periods and the rotation intedval, we need some adcitional notation. For all $f\le f$, we set $M(c,d):={\ensurelath{\{n \in \N \,\xolog c < k/n < d \tebt{ for somt nnteger $k$}\}}}$. Nktice that we do not assume herd thac $k$ and $n$ atz cootime. Obviousoy, $M(c,q)= \emptyset$ in and otly if $v=d$. Given $\rho \ik \R$ aid $S \subset \N$, we set $$\Lambva(\rho, S) = \begin{cases} \emptysat & \text{if $\ryo \notit \Q$}, \\ {\dbsufemzti{\{nq \,\colon q\ii S\}}} & \text{if $\rho=k/n$ with $k$ and $n$ coprime}. \end{vafvx}$$ The next thsorem wesalls Misiurewicz’s characterization of uhe ssts of periods for degrwe $1$ circle maps (see [@Mls; @ALM]). \[S9.5\] Lqt $F\in {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, and let $\operatortame{Rkg}(F) = [c,q]$. Rhfn there exist numbers $s_c,s_d \in \N_{\Sho}$ such that $\kptranorname{Per}(F)= \Lambdc(c,\operatorname{S\mnod{\togy\textup{sh}}}(s_c)) \zup M(c,b) \chp \Lambda(d,\operatorjame{S\mbjx{\tint\textup{sh}}}(f_d))$. Cpnversely, for all $c,d \in \R$ wuth $c \le d$ aud qll $s_c, s_d\in\N_{\Sho}$, thzre exists a map $F \in {\mathcal{L}}_1

{\ensuremath{\{z \in \C \,\colon |z| = 1\}}}$, ${\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$ the class all liftings of one. $F\in {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, $\operatorname{Rot}(F)$ the rotation set $F$ and, by [@Ito], is a non empty interval. To study the connection between the set of periods and rotation interval, we need some additional notation. For all $c\le d$, we set \in \,\colon < < d \text{ for some integer $k$}\}}}$. Notice that we do not assume here that $k$ $n$ are coprime. Obviously, $M(c,d)= \emptyset$ if and if $c=d$. Given $\rho \R$ and $S \subset \N$, set S) = \emptyset \text{if \notin \Q$}, \\ \,\colon q\in S\}}} & \text{if $\rho=k/n$ with $k$ and $n$ coprime}. \end{cases}$$ The next theorem recalls Misiurewicz’s of the periods for $1$ maps [@Mis; @ALM]). \[S9.5\] {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, and let $\operatorname{Rot}(F) = [c,d]$. numbers $s_c,s_d \in \N_{\Sho}$ such that $\operatorname{Per}(F)= \Lambda(c,\operatorname{S\mbox{\tiny\textup{sh}}}(s_c)) M(c,d) \cup Conversely, for all $c,d \in \R$ $c \le d$ and all $s_c, s_d\in\N_{\Sho}$, there a map $F \in {\mathcal{L}}_1

{\ensuremath{\{z \in \C \,\colon |z| = 1\}}}$, and leT ${\mathcal{L}}_1({\eNsureMatH{\maThBb{R}}})$ dEnotE the class of all LIftiNgs of continuous circle mAps of DeGRee oNE. IF $F\in {\mAthcal{L}}_1({\ENsUREmaTh{\MaThbB{R}}})$, $\OPeRatorNamE{Rot}(F)$ deNotes the roTatIoN set of $F$ and, by [@iTo], Is a compact Non Empty intervaL. To Study tHe ConNEctioN beTween The set OF perioDs and the rOtATion inTErval, we NEEd Some Additional notatioN. foR All $c\le d$, we set $M(c,D):={\ensurEmATh{\{N \IN \N \,\cOloN c < k/n < d \text{ fOr Some iNTeger $k$}\}}}$. NOTiCE THat WE do not assume hEre that $k$ and $N$ Are CoprimE. OBviOUsly, $M(c,D)= \emptYsET$ if And only if $c=d$. giveN $\rho \in \R$ anD $S \subsET \N$, we set $$\lAmbda(\rhO, S) = \begiN{caSes} \EmptYSeT & \tExt{If $\RHo \nOTiN \Q$}, \\ {\eNSurEmath{\{nq \,\cOlOn Q\in S\}}} & \tExt{iF $\RHO=K/n$ wiTh $k$ And $n$ CopriMe}. \end{cases}$$ The NexT theORem RecalLs MisIureWiCz’s chAracteRizatIoN of the sets of perIods For degree $1$ CirClE maPs (See [@MiS; @aLM]). \[S9.5\] LeT $F\iN {\maThcal{L}}_1({\eNsuremaTH{\maThBB{r}}})$, AnD let $\operatorname{RoT}(F) = [C,D]$. thEn there eXist nuMBeRs $S_C,s_d \in \N_{\ShO}$ sUch That $\OPEratoRnamE{peR}(F)= \Lambda(C,\operaTOrNaMe{S\mbox{\TiNy\textUp{Sh}}}(s_C)) \cuP M(c,d) \cUP \LamBda(d,\opEratornaMe{S\mbOX{\tiny\textup{sh}}}(s_D))$. conversely, for ALl $C,D \In \r$ With $C \le D$ and all $s_c, s_d\In\N_{\SHO}$, theRe exIStS a mAP $F \in {\mAthcaL{L}}_1

{\ensuremath{\{z \in \C \ ,\colon |z | = 1 \}} }$, a nd l et $ {\mathcal{L}}_ 1 ({\e nsuremath{\mathbb{R}}} )$ de no t e th e c lassof alll if t i ngs o fcon ti n uo us ci rcl e mapsof degreeone .If $F\in {\m a th cal{L}}_1( {\e nsuremath{\m ath bb{R}} }) $,$ \oper ato rname {Rot}( F )$ den otes thero t ations et of $ F $ a nd,by [@Ito], is a c o mp a ct non empty i nterva l. T o stu dythe connec ti on be t ween th e s e t ofp eriods and th e rotationi nte rval,we ne e d some addi ti o nal notation.Forall $c\le d$, w e set $M ( c,d):={ \ensur ema th{ \{n\ in \ N \ ,\ c olo n c <k /n< d \tex t{ f or so me i n t e g er $ k$} \}}} $. No tice that wedonota ssu me he re th at $ k$ and$n$ ar e cop ri me. Obviously,$M(c ,d)= \emp tys et $ i fand o n ly if$c= d$. Given$\rho \ i n \ R$ a n d$S \subset \N$, we s e t $ $\Lambda (\rho, S) = \begin{c as es} \ empty set & \t ext{if $\ rh o \noti n\Q$},\\ { \ensu r emat h{\{nq \,\colo n q\i n S\}}} & \text { if $\rho=k/n$ wi t h $ k $ an d $ n$ coprime} . \e n d{ca ses} $ $ Th e next theo re m r e calls Misiurewicz’s c haract eriza tion of the s ets of per i o d s for de gree $1 $ circle maps ( see [ @Mis; @ALM ] ). \[S9 .5\]Let $F\i n {\mathc a l {L}}_1({ \en sur ema th{ \ m at hbb{R}}})$, a n d let $ \operat orn ame{Rot }(F ) = [c ,d] $. Then the re exist n um be rs $s _c,s_ d \in \N_ {\ Sho }$ su ch th a t $\op erato rnam e{ Pe r }(F )= \Lam b da ( c ,\op er at orna me{ S\ mbox{ \tin y \te xtup{sh }}}(s_c)) \c u p M( c, d) \cup \ Lambda(d,\ope ra torname{S\ mb ox{ \tiny\ t e xtup{sh} }}(s_d))$. Conversely,f or all$c, d \in \R$ with $c\le d$ an d a l l $s_c , s_d\ in\N_ {\ Sho } $ , the r e e xis ts a map $F\ i n { \math ca l{L} }_1

{\ensuremath{\{z \in_\C \,\colon_|z| = 1\}}}$, and_let ${\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$_denote_the class_of_all liftings of_continuous circle maps_of degree one. If_$F\in {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$, $\operatorname{Rot}(F)$ denotes_the_rotation set of $F$ and, by [@Ito], is a compact non empty interval. To study_the_connection between_the_set_of periods and the rotation_interval, we need some additional_notation. For_all $c\le d$, we set $M(c,d):={\ensuremath{\{n \in \N_\,\colon_c < k/n_< d \text{ for some integer $k$}\}}}$. Notice that_we do not assume here that_$k$ and $n$_are_coprime._Obviously, $M(c,d)= \emptyset$ if_and only if $c=d$. Given $\rho_\in \R$ and $S \subset \N$,_we set $$\Lambda(\rho, S) = \begin{cases} _ \emptyset _ &_\text{if $\rho_\notin \Q$}, \\ _ {\ensuremath{\{nq \,\colon_q\in S\}}}_& \text{if $\rho=k/n$_with $k$ and $n$ coprime}. \end{cases}$$ The next_theorem recalls Misiurewicz’s_characterization of the sets of periods_for_degree $1$ circle_maps_(see_[@Mis; @ALM]). \[S9.5\]_Let $F\in {\mathcal{L}}_1({\ensuremath{\mathbb{R}}})$,_and_let $\operatorname{Rot}(F)_=_[c,d]$. Then there exist numbers $s_c,s_d_\in_\N_{\Sho}$ such that $\operatorname{Per}(F)= \Lambda(c,\operatorname{S\mbox{\tiny\textup{sh}}}(s_c)) \cup M(c,d)_\cup \Lambda(d,\operatorname{S\mbox{\tiny\textup{sh}}}(s_d))$. Conversely, for_all_$c,d \in \R$ with_$c \le d$ and all_$s_c, s_d\in\N_{\Sho}$, there exists a map_$F \in_{\mathcal{L}}_1

range of the spin diffusion length $\lambda_m$, which is clearly of an advantage for designing spin-based, non-volatile nanoelectronic devices.\ In Sec. \[sec1\] we discuss the mathematical details of the spin-spiral magnetoelectric coupling, followed by its implementation into the equations of motion for the magnetization dynamics in Sec. \[sec2\]. In Sec. \[sec3\] results of numerical simulations are presented and discussed showing to which extent the spin-spiral magnetoelectric coupling can allow for the electric field control of the magnetization in FE/FM composites. Ways to enhance the effects are discussed and brief conclusions are made in Sec. \[sec4\]. Theoretically, the above magnon accumulation scenario maybe viewed as follows: $$\mathcal{F}_{sd}=J_{sd}\frac{M}{M_{s}}{\mbox{\boldmath$\mathrm{s}$}}\cdot {{\mbox{\boldmath$\mathrm{m}$}}}, \label{eq:sd}$$ Within the Stoner mean-field theory [@Soulen; @R.J:1998] the spin polarization $\eta$ of the electron density in transition FM metals is usually less than 1, $${\mbox{\boldmath$\mathrm{s}$}}={\mbox{\boldmath$\mathrm{s}$}}_{\parallel}+{\mbox{\boldmath$\mathrm{s}$}}_{\perp}$$ where ${\mbox{\boldmath$\mathrm{s}$}}_{\parallel}$ represents the spin density whose direction follows adiabatically the intrinsic magnetization ${\mbox{\boldmath$\mathrm{M}$}}$ at an instantaneous time t. ${\mbox{\boldmath$\mathrm{s}$}}_{\perp}$ describes the transverse deviation from ${\mbox{\boldmath$\mathrm{M}$}}$. Given that the steady-state charge accumulation entails much higher energy processes than spin excitations, $$\begin{aligned} &&\frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\partial t}{\mbox{\boldmath$\mathrm{m}$}}+{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}\frac{\partial{\mbox{\boldmath$\mathrm{m}$}}}{\partial t}+\frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\perp}}{\partial t} -D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}-D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\perp} \nonumber \\ &&= -\frac{{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\tau_{sf}}-\frac{{\mbox{\boldmath$\mathrm

range of the spin diffusion length $ \lambda_m$, which is intelligibly of an advantage for design spin - based, non - volatile nanoelectronic devices.\ In Sec. \[sec1\ ] we hash out the mathematical details of the tailspin - spiral magnetoelectric coupling, followed by its execution into the equation of motion for the magnetization moral force in Sec. \[sec2\ ]. In Sec. \[sec3\ ] results of numerical simulations are presented and discourse showing to which extent the spin - spiral magnetoelectric coupling can allow for the electric battlefield control of the magnetization in FE / FM composites. Ways to enhance the effect are discussed and brief conclusions are made in Sec. \[sec4\ ]. Theoretically, the above magnon accumulation scenario possibly viewed as follow: $ $ \mathcal{F}_{sd}=J_{sd}\frac{M}{M_{s}}{\mbox{\boldmath$\mathrm{s}$}}\cdot { { \mbox{\boldmath$\mathrm{m}$ } } }, \label{eq: sd}$$ Within the Stoner mean - field theory [ @Soulen; @R.J:1998 ] the tailspin polarization $ \eta$ of the electron density in transition FM metals is usually less than 1, $ $ { \mbox{\boldmath$\mathrm{s}$}}={\mbox{\boldmath$\mathrm{s}$}}_{\parallel}+{\mbox{\boldmath$\mathrm{s}$}}_{\perp}$$ where $ { \mbox{\boldmath$\mathrm{s}$}}_{\parallel}$ represents the spin concentration whose guidance follows adiabatically the intrinsic magnetization $ { \mbox{\boldmath$\mathrm{M}$}}$ at an instantaneous time t. $ { \mbox{\boldmath$\mathrm{s}$}}_{\perp}$ describes the transverse deviation from $ { \mbox{\boldmath$\mathrm{M}$}}$. Given that the steady - state charge accretion entails much high energy processes than spin excitations, $ $ \begin{aligned } & & \frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\partial t}{\mbox{\boldmath$\mathrm{m}$}}+{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}\frac{\partial{\mbox{\boldmath$\mathrm{m}$}}}{\partial t}+\frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\perp}}{\partial deoxythymidine monophosphate } -D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}-D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\perp } \nonumber \\ & & = -\frac{{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\tau_{sf}}-\frac{{\mbox{\boldmath$\mathrm

rajge of the spin diffusiok length $\lambda_m$, which ms cleadly of av advantage for designing spmn-bawed, nin-volatile nanoelectrovic devicvs.\ In Sec. \[wec1\] xe discuss the mefhematigcl defwils if the spin-spital magnetoenectric couplitg, flllowed by its implementation into ehe equstlons of motion for ehe jagnetization dynamics in Sec. \[sec2\]. Jn Sec. \[vec3\] results og numerical simulations arf prfsented and discusded showing to ryich extent ghe spin-sppxal magnetoglectric coupling can allow for ghe ekectric figls fmntrol of tie maggetization ik FE/FM womposiyes. Ways to enmance thw effects are discussxd and brief conclusyons are kabe in Sec. \[sec4\]. Theoretixaoly, tve atove nagvon arcujulatiln acenario mzybe viewed as follows: $$\mathcal{G}_{sq}=H_{sd}\frac{M}{M_{s}}{\mbos{\boldmwtr$\mathrm{s}$}}\cdot {{\mbox{\boldmath$\mathrm{m}$}}}, \label{eq:vd}$$ Sithin the Stoner mean-fueld theory [@Soulen; @R.J:1998] the spin polarization $\eta$ of the electron density in tranvitioi WM nebals us usually less than 1, $${\mbox{\boldmath$\mathrm{s}$}}={\mbox{\bojsmstm$\mathrm{s}$}}_{\parallel}+{\ibox{\boldmatn$\mwtntm{s}$}}_{\perp}$$ where ${\obox{\boldmzth$\mathrm{s}$}}_{\parallel}$ represgnts tye spin dtnsitu whose direction follows aeiabatically rhe intrinsic magnztization ${\mbux{\bokdmatn$\mathrm{M}$}}$ at an instantaueous fime t. ${\mbox{\holdmath$\mzghrm{s}$}}_{\perp}$ descrices tve transverse deviation frjm ${\mbox{\boodmach$\mathrm{O}$}}$. Gifen thwt the stewdy-state charge accumulahion gntailv much higjer energy processes than spin xecitations, $$\bebit{alpgned} &&\frac{\'artiak{\mbox{\boldmath$\iathrm{s}$}}_{\parallek}}{\partiaj t}{\mbux{\boldmath$\jathrm{m}$}}+{\kbox{\boldmatr$\mathrm{s}$}}_{\paralngl}\frac{\partial{\kbox{\boldiath$\nathem{m}$}}}{\partkxl t}+\frac{\partiak{\mbox{\boldmath$\mathrm{w}$}}_{\perp}}{\partial t} -D_{0}\nanla^{2}_{z}{\mckx{\boldmath$\mathrn{s}$}}_{\pqrallel}-D_{0}\nabla^{2}_{z}{\mnox{\colqmwti$\mathwk{s}$}}_{\perp} \nonumter \\ &&= -\frxv{{\mbox{\coldmath$\matmrm{r}$}}_{\parsllel}}{\tau_{sf}}-\frac{{\mbox{\bondmafh$\mathrm

range of the spin diffusion length $\lambda_m$, clearly an advantage designing spin-based, non-volatile we the mathematical details the spin-spiral magnetoelectric followed by its implementation into the of motion for the magnetization dynamics in Sec. \[sec2\]. In Sec. \[sec3\] results numerical simulations are presented and discussed showing to which extent the spin-spiral magnetoelectric can for electric control of the magnetization in FE/FM composites. Ways to enhance the effects are discussed and brief are made in Sec. \[sec4\]. Theoretically, the above accumulation scenario maybe viewed follows: $$\mathcal{F}_{sd}=J_{sd}\frac{M}{M_{s}}{\mbox{\boldmath$\mathrm{s}$}}\cdot {{\mbox{\boldmath$\mathrm{m}$}}}, \label{eq:sd}$$ Within Stoner theory [@Soulen; the polarization of the electron in transition FM metals is usually less than 1, $${\mbox{\boldmath$\mathrm{s}$}}={\mbox{\boldmath$\mathrm{s}$}}_{\parallel}+{\mbox{\boldmath$\mathrm{s}$}}_{\perp}$$ where ${\mbox{\boldmath$\mathrm{s}$}}_{\parallel}$ represents the spin density whose follows adiabatically magnetization ${\mbox{\boldmath$\mathrm{M}$}}$ an time ${\mbox{\boldmath$\mathrm{s}$}}_{\perp}$ describes the from ${\mbox{\boldmath$\mathrm{M}$}}$. Given that the steady-state much higher energy processes than spin excitations, $$\begin{aligned} t}{\mbox{\boldmath$\mathrm{m}$}}+{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}\frac{\partial{\mbox{\boldmath$\mathrm{m}$}}}{\partial t}+\frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\perp}}{\partial -D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}-D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\perp} \nonumber \\ &&= -\frac{{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\tau_{sf}}-\frac{{\mbox{\boldmath$\mathrm

range of the spin diffusion leNgth $\lambda_M$, whicH is CleArLy of An adVantage for desiGNing Spin-based, non-volatile naNoeleCtROnic DEvIces.\ IN Sec. \[sec1\] WE dISCusS tHe MatHeMAtIcal dEtaIls of thE spin-spiraL maGnEtoelectric cOUpLing, followEd bY its implemenTatIon intO tHe eQUatioNs oF motiOn for tHE magneTization dYnAMics in sEc. \[sec2\]. In sEC. \[sEc3\] reSults of numerical sIMuLAtions are preseNted anD dIScUSSed ShoWing to whicH eXtent THe spin-sPIrAL MAgnEToelectric couPling can allOW foR the elEcTriC Field cOntroL oF The MagnetizatiOn in fE/FM compoSites. WAYs to enhANce the eFfects Are DisCussED aNd BriEf COncLUsIonS Are Made in SeC. \[sEc4\]. theorEticALLY, The aBovE magNon acCumulation sceNarIo maYBe vIewed As folLows: $$\MaThcal{f}_{sd}=J_{sd}\Frac{M}{m_{s}}{\Mbox{\boldmath$\matHrm{s}$}}\Cdot {{\mbox{\bOldMaTh$\mAtHrm{m}$}}}, \lABel{eq:sD}$$ WiThiN the StoNer mean-FIelD tHEORy [@soulen; @R.J:1998] the spin polArIZAtIon $\eta$ of The eleCTrOn DEnsity in TrAnsItioN fm metaLs is USuAlly less Than 1, $${\mbOX{\bOlDmath$\maThRm{s}$}}={\mboX{\bOldMatH$\mathRM{s}$}}_{\paRallel}+{\Mbox{\boldMath$\mAThrm{s}$}}_{\perp}$$ where ${\MBox{\boldmath$\maTHrM{S}$}}_{\PaRAlleL}$ rePresents the Spin DEnsiTy whOSe DirECtion FolloWs ADiABatically the intrinsIc MagnetIzatiOn ${\mbox{\boldmatH$\mathrm{M}$}}$ at AN INstantanEous TImE T. ${\mbox{\boldmath$\mAthrm{S}$}}_{\perp}$ descrIBes the trAnsveRse deviaTion from ${\mBOX{\boldmatH$\maThrM{M}$}}$. GIveN THaT the steady-staTE CharGe AccumulAtiOn entaiLs mUch HigHer EnErgy proceSses than SpIn ExCiTatIons, $$\bEGin{alignEd} &&\FraC{\pArtIal{\mbOX{\boldmAth$\maThrm{S}$}}_{\pArALleL}}{\partiaL T}{\mBOX{\bolDmAtH$\matHrm{M}$}}+{\mBox{\boLdmaTH$\maThrm{s}$}}_{\paRallel}\fraC{\paRTial{\MbOx{\BoldmatH$\mathrm{m}$}}}{\partiAl T}+\frac{\partiAl{\MboX{\boldmATH$\mathrm{s}$}}_{\Perp}}{\partial t} -D_{0}\nabla^{2}_{z}{\mbox{\BOldmath$\MatHrm{s}$}}_{\pAralLel}-D_{0}\nabla^{2}_{Z}{\mbOx{\boldMatH$\Mathrm{S}$}}_{\perp} \nOnumbEr \\ &&= -\FraC{{\MBox{\boLDMaTh$\mAtHrm{s}$}}_{\parallEL}}{\Tau_{Sf}}-\fraC{{\mBox{\bOldmath$\Mathrm

range of the spin diffusi on length$\lam bda _m$ ,whic h is clearly of an adva ntage for designing sp in-ba se d , no n -v olati le nano e le c t ron ic d evi ce s .\ In S ec. \[sec1 \] we disc uss t he mathemati c al details o f t he spin-spir almagnet oe lec t ric c oup ling, follo w ed byits imple me n tation into th e eq uati ons of motion for th e magnetization dynam ic s i n Sec . \ [sec2\]. I nSec.\ [sec3\] re s u l tso f numerical s imulationsa represen te d a n d disc ussed s h owi ng to which ext ent the s pin-sp i ral mag n etoelec tric c oup lin g ca n a ll owfo r th e e lec t ric field c on tr ol of the m a g neti zat ionin FE /FM composite s.Ways toenhan ce th e ef fe cts a re dis cusse dand brief concl usio ns are ma dein Se c. \[se c 4\]. The ore tically , the a b ove m a g n on accumulation scen ar i o m aybe vie wed as fo ll o ws: $$\m at hca l{F} _ { sd}=J _{sd } \f rac{M}{M _{s}}{ \ mb ox {\boldm at h$\mat hr m{s }$} }\cdo t {{\ mbox{\ boldmath $\mat h rm{m}$}}}, \la b el{eq:sd}$$ W i th i n t h e St one r mean-fiel d th e ory[@So u le n;@ R.J:1 998]th e s p in polarization $\e ta $ of t he el ectron densit y in trans i t i on FM me tals is usually less t han 1 , $${\mbox { \boldmat h$\ma thrm{s}$ }}={\mbox { \ boldmath $\m ath rm{ s}$ } } _{ \parallel}+{\ m b ox{\ bo ldmath$ \ma thrm{s} $}} _{\ per p}$ $where ${\ mbox{\bo ld ma th $\ mat hrm{s } $}}_{\pa ra lle l} $ r epres e nts th e spi n de ns it y wh ose dir e ct i o n fo ll ow s ad iab at icall y th e in trinsic magnetiz ati o n ${ \m bo x{\bold math$\mathrm{ M} $}}$ at an i nst antane o u s time t . ${\mbox{\boldmath$\ma t hrm{s}$ }}_ {\per p}$describes th e tran sve r se dev iation from $ {\m b o x{\bo l d ma th$ \m athrm{M}$} } $ . G iventh at t he stea dy-state charge ac c umu lation entail s m uchh i gh ere ne r gypr o ces s e s than spin exc itations,$$ \ be gin{aligne d } & &\ frac{\p artial{ \mbox { \boldma th$\mathr m{s}$}}_{ \p aral l e l}} {\partialt}{\mbox {\boldmat h $\mat h rm {m}$} }+{ \mbox{ \b old math$ \mathr m {s} $}}_{ \paral le l}\fra c{\pa rt ial{\mbo x{\boldmath$\mathrm{m}$ }}}{\p artia l t }+\frac{\ par t ial {\mbox{\b oldm ath$\mathr m{s }$} }_{\p erp } }{\pa rtia l t } - D _{0}\ nabl a ^{2}_{z}{ \ mb ox{ \ b ol dmath$\math r m { s}$ }}_{\ par a llel}- D_{0 }\nabla^{2}_{z}{\ m box{\boldmath$ \mat h r m{s }$} } _{\p er p} \nonumber \ \ & &= - \frac{{\ mb ox{\boldmat h$\mathr m{ s }$}}_ {\para llel}} {\tau_{ s f }} - \frac{ {\mb ox{ \boldmath $\m at h rm

range_of the_spin diffusion length $\lambda_m$,_which is_clearly_of an_advantage_for designing spin-based,_non-volatile nanoelectronic devices.\ In_Sec. \[sec1\] we discuss_the mathematical details_of_the spin-spiral magnetoelectric coupling, followed by its implementation into the equations of motion for_the_magnetization dynamics_in_Sec._\[sec2\]. In Sec. \[sec3\] results_of numerical simulations are presented_and discussed_showing to which extent the spin-spiral magnetoelectric coupling_can_allow for the_electric field control of the magnetization in FE/FM composites._Ways to enhance the effects are_discussed and brief_conclusions_are_made in Sec. \[sec4\]. Theoretically,_the above magnon accumulation scenario maybe_viewed as follows: $$\mathcal{F}_{sd}=J_{sd}\frac{M}{M_{s}}{\mbox{\boldmath$\mathrm{s}$}}\cdot {{\mbox{\boldmath$\mathrm{m}$}}}, \label{eq:sd}$$ Within_the Stoner mean-field theory [@Soulen; @R.J:1998] the_spin polarization $\eta$ of the electron_density in transition FM metals_is usually_less than 1, $${\mbox{\boldmath$\mathrm{s}$}}={\mbox{\boldmath$\mathrm{s}$}}_{\parallel}+{\mbox{\boldmath$\mathrm{s}$}}_{\perp}$$ where_${\mbox{\boldmath$\mathrm{s}$}}_{\parallel}$ represents the_spin density_whose direction follows_adiabatically the intrinsic magnetization ${\mbox{\boldmath$\mathrm{M}$}}$ at_an instantaneous time_t. ${\mbox{\boldmath$\mathrm{s}$}}_{\perp}$ describes the transverse deviation_from_${\mbox{\boldmath$\mathrm{M}$}}$. Given that_the_steady-state_charge accumulation_entails much higher_energy_processes than_spin_excitations, $$\begin{aligned} &&\frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\partial t}{\mbox{\boldmath$\mathrm{m}$}}+{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}\frac{\partial{\mbox{\boldmath$\mathrm{m}$}}}{\partial t}+\frac{\partial{\mbox{\boldmath$\mathrm{s}$}}_{\perp}}{\partial t} -D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}-D_{0}\nabla^{2}_{z}{\mbox{\boldmath$\mathrm{s}$}}_{\perp}_\nonumber_\\ &&= -\frac{{\mbox{\boldmath$\mathrm{s}$}}_{\parallel}}{\tau_{sf}}-\frac{{\mbox{\boldmath$\mathrm

ge 1$, every point $z$ in $J(f)$ and every pair of points $x$ and $x'$ in the same pull-back $W$ of $B(z, \rho_0)$ by $f^n$, we have $$\begin{aligned} |S_n(\varphi)(x) - S_n(\varphi)(x')| & = \left| \sum_{i=0}^{n-1} \varphi(f^i(x))- \sum_{i=0}^{n-1}\varphi(f^i(x')) \right| \\ & \le \sum_{i=0}^{n-1} C_*\operatorname{diam}(f^i(W))^{\alpha} \\ & \le \sum_{i=0}^{n-1}C_* C_0^{\alpha} (n-i)^{- \beta \alpha} \\ & \le C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty} m^{- \beta\alpha}.\end{aligned}$$ This proves the lemma with constants $\rho_1 = \rho_0$ and $C_1 = C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty} m^{-\beta\alpha}$. \[l:maximizing branches\] Let $f$ be a map in ${\mathscr{A}}$, let $\varphi: J(f)\rightarrow {\mathbb{R}}$ be a continuous function, and let $\nu$ be an invariant and ergodic Borel probability measure supported on $J(f)$. Then there exists a Borel subset $X'$ of $J(f)$ of full measure with respect to $\nu$, such that the following holds. For every $D' > 0$, and every point $x$ of $X'$ there exist a strictly increasing sequence positive integers $(n_l)_{l = 1}^{+ \infty}$ such that for every $l$ we can choose a point $x_l$ in $f^{-n_l}(x)$ so that: 1. $x_{l + 1}$ is in $f^{-(n_{l + 1} - n_l)}(x_l)$; 2. $S_{n_l}(\varphi)(x_l

ge 1 $, every point   $ z$ in   $ J(f)$ and every pair of points   $ x$ and   $ x'$ in the same wrench - binding   $ W$ of   $ B(z, \rho_0)$ by   $ f^n$, we have $ $ \begin{aligned } |S_n(\varphi)(x) - S_n(\varphi)(x')| & = \left| \sum_{i=0}^{n-1 } \varphi(f^i(x))- \sum_{i=0}^{n-1}\varphi(f^i(x') ) \right| \\ & \le \sum_{i=0}^{n-1 } C_*\operatorname{diam}(f^i(W))^{\alpha } \\ & \le \sum_{i=0}^{n-1}C _ * C_0^{\alpha } (n - i)^{- \beta \alpha } \\ & \le C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty } m^{- \beta\alpha}.\end{aligned}$$ This proves the lemma with constants   $ \rho_1 = \rho_0 $ and   $ C_1 = C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty } m^{-\beta\alpha}$. \[l: maximize branches\ ] Let   $ f$ be a map in   $ { \mathscr{A}}$, permit $ \varphi: J(f)\rightarrow { \mathbb{R}}$ be a continuous function, and let   $ \nu$ be an changeless and ergodic Borel probability standard supported on   $ J(f)$. Then there exist a Borel subset   $ X'$ of   $ J(f)$ of full measure with obedience to   $ \nu$, such that the following holds. For every   $ D' > 0 $, and every point   $ x$ of   $ X'$ there exist a strictly increasing succession positive integers $ (n_l)_{l = 1}^{+ \infty}$ such that for every   $ l$ we can choose a point   $ x_l$ in   $ f^{-n_l}(x)$ so that: 1. $ x_{l + 1}$ is in   $ f^{-(n_{l + 1 } - n_l)}(x_l)$; 2. $ S_{n_l}(\varphi)(x_l

ge 1$, every point $z$ in $J(f)$ and eyery pair of poiurs $x$ anv $x'$ in tge same oull-back $W$ of $B(z, \rho_0)$ by $f^n$, we heve $$\vegin{qligned} |S_n(\varphi)(x) - S_n(\vafphi)(x')| & = \lefn| \sum_{i=0}^{n-1} \vqrphm(f^i(x))- \sum_{i=0}^{n-1}\varphi(h^j(x')) \righb| \\ & \le \shl_{i=0}^{n-1} E_*\o'eratorname{diam}(n^i(W))^{\alpha} \\ & \la \sum_{i=0}^{n-1}C_* C_0^{\alpha} (n-k)^{- \yeta \alpha} \\ & \le C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty} m^{- \feta\alpna}.\fnd{aligned}$$ Thif prpdes fhe lemma with constants $\rho_1 = \rho_0$ ahd $C_1 = C_*C_0^{\elpha}\sum_{m = 1}^{+ \infyy} m^{-\beta\alpha}$. \[l:maximizing bgancjes\] Let $f$ be a map ln ${\mathscr{A}}$, oet $\dqrphi: J(f)\righgarrow {\mathbb{R}}$ be a cohtinuous function, and let $\nu$ be xn inrariant and eegofhc Borel pribabijity measure supportad on $J(f)$. Then there exlsts e Boeel subset $X'$ of $J(f)$ of fnll measure with reskect to $\nu$, vueh that the following hilds. Xor averh $E' > 0$, ans xvedy poijt $x$ of $X'$ there exist a steictly increasing stquqbce positive jntegews $(n_l)_{l = 1}^{+ \infty}$ such that for every $l$ we cat cgoose a point $x_l$ in $f^{-n_l}(x)$ wo that: 1. $x_{l + 1}$ is in $f^{-(n_{p + 1} - n_l)}(x_l)$; 2. $S_{n_l}(\varphi)(x_l

ge 1$, every point $z$ in $J(f)$ pair points $x$ $x'$ in the \rho_0)$ $f^n$, we have |S_n(\varphi)(x) - S_n(\varphi)(x')| = \left| \sum_{i=0}^{n-1} \varphi(f^i(x))- \sum_{i=0}^{n-1}\varphi(f^i(x')) \right| & \le \sum_{i=0}^{n-1} C_*\operatorname{diam}(f^i(W))^{\alpha} \\ & \le \sum_{i=0}^{n-1}C_* C_0^{\alpha} (n-i)^{- \beta \alpha} \\ \le C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty} m^{- \beta\alpha}.\end{aligned}$$ This proves the lemma with constants = and = = 1}^{+ \infty} m^{-\beta\alpha}$. \[l:maximizing branches\] Let $f$ be a map in ${\mathscr{A}}$, let $\varphi: J(f)\rightarrow be a continuous function, and let $\nu$ be invariant and ergodic Borel measure supported on $J(f)$. Then exists Borel subset of of measure with respect $\nu$, such that the following holds. For every $D' > 0$, and every point $x$ of $X'$ exist a sequence positive $(n_l)_{l 1}^{+ such that for we can choose a point $x_l$ that: 1. $x_{l + 1}$ is in $f^{-(n_{l 1} - 2. $S_{n_l}(\varphi)(x_l

ge 1$, every point $z$ in $J(f)$ and every Pair of poinTs $x$ anD $x'$ iN thE sAme pUll-bAck $W$ of $B(z, \rho_0)$ by $f^N$, We haVe $$\begin{aligned} |S_n(\varphi)(X) - S_n(\vaRpHI)(x')| & = \leFT| \sUm_{i=0}^{n-1} \vArphi(f^i(X))- \SuM_{I=0}^{N-1}\vaRpHi(F^i(x')) \RiGHt| \\ & \Le \sum_{I=0}^{n-1} C_*\OperatoRname{diam}(f^I(W))^{\aLpHa} \\ & \le \sum_{i=0}^{n-1}C_* C_0^{\aLPhA} (n-i)^{- \beta \alpHa} \\ & \lE C_*C_0^{\alpha}\sum_{m = 1}^{+ \InfTy} m^{- \betA\aLphA}.\End{alIgnEd}$$ ThiS proveS The lemMa with conStANts $\rho_1 = \RHo_0$ and $C_1 = C_*c_0^{\ALpHa}\suM_{m = 1}^{+ \infty} m^{-\beta\alpha}$. \[L:MaXImizing brancheS\] Let $f$ bE a MAp IN ${\MatHscR{A}}$, let $\varphI: J(F)\righTArrow {\maTHbB{r}}$ BE a cONtinuous functIon, and let $\nu$ BE an InvariAnT anD ErgodiC BoreL pRObaBility measuRe suPported on $j(f)$. Then THere exiSTs a BoreL subseT $X'$ oF $J(f)$ Of fuLL mEaSurE wITh rESpEct TO $\nu$, Such that ThE fOllowIng hOLDS. for eVerY $D' > 0$, anD everY point $x$ of $X'$ theRe eXist A StrIctly IncreAsinG sEquenCe posiTive iNtEgers $(n_l)_{l = 1}^{+ \infty}$ suCh thAt for everY $l$ wE cAn cHoOse a pOInt $x_l$ iN $f^{-n_L}(x)$ sO that: 1. $x_{l + 1}$ Is in $f^{-(n_{l + 1} - N_L)}(x_l)$; 2. $s_{n_L}(\VARpHi)(x_l

ge 1$, every point $z$ in$J(f)$ and ever y p air o f po ints  $x$ and $x'$i n th e same pull-back $W$ o f $B( z, \rho _ 0) $ by$f^n$,w eh a ve$$ \b egi n{ a li gned} |S _n(\var phi)(x) -S_n (\ varphi)(x')| &= \left| \ sum _{i=0}^{n-1} \v arphi( f^ i(x ) )- \s um_ {i=0} ^{n-1} \ varphi (f^i(x')) \ r ight|\ \ & \le \ su m_{i =0}^{n-1} C_*\ope r at o rname{diam}(f^ i(W))^ {\ a lp h a } \ \ & \le \sum_ {i =0}^{ n -1}C_*C _0 ^ { \ alp h a} (n-i)^{- \ beta \alpha } \\ & \le C _*C _ 0^{\al pha}\ su m _{m = 1}^{+ \i nfty } m^{- \b eta\al p ha}.\en d {aligne d}$$ T his pr oves th elem ma wit h c ons t ant s $\rho_ 1=\rho_ 0$ a n d $ C_1= C _*C_ 0^{\a lpha}\sum_{m= 1 }^{+ \in fty}m^{-\ beta \a lpha} $. \[ l:max im izing branches\ ] Le t $f$ bea m ap in  $ {\mat h scr{A} }$, le t $\var phi: J( f )\r ig h t a rr ow {\mathbb{R}}$ b ea co ntinuous funct i on ,a nd let $ \n u$be a n invar iant an d ergodi c Bore l p ro babilit ymeasur esup por ted o n  $J( f)$. T hen ther e exi s ts a Borel sub s et $X'$ of $J ( f) $ of full me asure withresp e ct t o $\ n u$ , s u ch th at th ef ol l owing holds. For ev er y $D'> 0$, and every po int $x$ of $ X '$ there exi s ta strictly incr easin g sequence positive inte gers $(n _l)_{l =1 } ^{+ \inf ty} $ s uch th a t f or every $l$w e can c hoose a po int $x_ l$in$f^ {-n _l }(x)$ sothat: 1 . $ x_ {l +1}$ i s in $f^{ -( n_{ l+ 1 } - n _ l)}(x_ l)$; 2. $ S_ { n_l }(\varp h i) ( x _l

ge 1$,_every point $z$_in $J(f)$ and every pair_of points $x$_and $x'$_in the_same_pull-back $W$ of $B(z, \rho_0)$_by $f^n$, we have_$$\begin{aligned} |S_n(\varphi)(x) - S_n(\varphi)(x')| & = \left|_\sum_{i=0}^{n-1} \varphi(f^i(x))- \sum_{i=0}^{n-1}\varphi(f^i(x'))_\right| \\_& \le \sum_{i=0}^{n-1} C_*\operatorname{diam}(f^i(W))^{\alpha} \\ & \le \sum_{i=0}^{n-1}C_* C_0^{\alpha} (n-i)^{- \beta \alpha} \\ & \le C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty} m^{-_\beta\alpha}.\end{aligned}$$_This proves_the_lemma_with constants $\rho_1 = \rho_0$ and $C_1_= C_*C_0^{\alpha}\sum_{m = 1}^{+ \infty}_m^{-\beta\alpha}$. \[l:maximizing branches\]_Let $f$ be a map in ${\mathscr{A}}$, let $\varphi: J(f)\rightarrow_{\mathbb{R}}$_be a continuous_function, and let $\nu$ be an invariant and ergodic Borel_probability measure supported on $J(f)$. Then there_exists a Borel_subset $X'$_of $J(f)$_of full measure with_respect to $\nu$, such that the following_holds. For every $D' > 0$, and_every point $x$ of $X'$ there exist a strictly_increasing sequence positive integers $(n_l)_{l =_1}^{+ \infty}$ such that for_every $l$ we_can choose a point $x_l$ in $f^{-n_l}(x)$_so that: 1. _$x_{l +_1}$ is in $f^{-(n_{l_+ 1} - n_l)}(x_l)$; 2. $S_{n_l}(\varphi)(x_l

]{}](http://arXiv.org/abs/hep-th/9907219). Tomasiello. D. L. Jafferis and A. Tomasiello, “[A simple class of N=3 gauge/gravity duals]{},” [[arXiv:0808.0864]{}](http://arXiv.org/abs/arXiv:0808.0864). A. Hanany and A. Zaffaroni, “[Tilings, Chern-Simons Theories and M2 Branes]{},” [*JHEP*]{} [**0810**]{} (2008) 111, [[ 0808.1244]{}](http://arXiv.org/abs/0808.1244). D. Martelli and J. Sparks, “[Notes on toric Sasaki-Einstein seven-manifolds and $AdS_4/CFT_3$]{},” [[arXiv:0808.0904]{}](http://arXiv.org/abs/arXiv:0808.0904). A. Hanany, D. Vegh, and A. Zaffaroni, “[Brane Tilings and M2 Branes]{},” [ *JHEP*]{} [**0903**]{} (2009) 012, [[ 0809.1440]{}](http://arXiv.org/abs/0809.1440). D. Martelli and J. Sparks, “[AdS4/CFT3 duals from M2-branes at hypersurface singularities and their deformations]{},” [*JHEP*]{} [**12**]{} (2009) 017, [[0909.2036]{}](http://arXiv.org/abs/0909.2036). S. Franco, I. R. Klebanov, and D. Rodriguez-Gomez, “[M2-branes on Orbifolds of the Cone over Q\*\*1,1,1]{},” [*JHEP*]{} [**0908**]{} (2009) 033, [[0903.3231]{}](http://arXiv.org/abs/0903.3231). F. Benini, C. Closset, and S. Cremonesi, “[Chiral flavors and M2-branes at toric CY4 singularities]{},” [*JHEP*]{} [**1002**]{} (2010) 036, [[0911.4127]{}](http://arXiv.

] { } ] (http://arXiv.org/abs/hep-th/9907219). Tomasiello. D.   L. Jafferis and A.   Tomasiello, “ [ A simple class of N=3 gauge / gravity duals ] { }, ” [ [ arXiv:0808.0864]{}](http://arXiv.org / abs / arXiv:0808.0864). A.   Hanany and A.   Zaffaroni, “ [ tiling, Chern - Simons Theories and M2 Branes ] { }, ” [ * JHEP * ] { } [ * * 0810 * * ] { } (2008) 111, [ [ 0808.1244]{}](http://arXiv.org / abs/0808.1244). D.   Martelli and J.   Sparks, “ [ eminence on toric Sasaki - Einstein seven - manifolds and $ AdS_4 / CFT_3 $ ] { }, ” [ [ arXiv:0808.0904]{}](http://arXiv.org / abs / arXiv:0808.0904). A.   Hanany, D.   Vegh, and A.   Zaffaroni, “ [ Brane Tilings and M2 Branes ] { }, ” [ * JHEP * ] { } [ * * 0903 * * ] { } (2009) 012, [ [ 0809.1440]{}](http://arXiv.org / abs/0809.1440). D.   Martelli and J.   Sparks, “ [ AdS4 / CFT3 duals from M2 - branes at hypersurface singularity and their deformation ] { }, ” [ * JHEP * ] { } [ * * 12 * * ] { } (2009) 017, [ [ 0909.2036]{}](http://arXiv.org / abs/0909.2036). S.   Franco, I.   R. Klebanov, and D.   Rodriguez - Gomez, “ [ M2 - branes on Orbifolds of the Cone over Q\*\*1,1,1 ] { }, ” [ * JHEP * ] { } [ * * 0908 * * ] { } (2009) 033, [ [ 0903.3231]{}](http://arXiv.org / abs/0903.3231). F.   Benini, C.   Closset, and S.   Cremonesi, “ [ Chiral flavors and M2 - branes at toric CY4 singularities ] { }, ” [ * JHEP * ] { } [ * * 1002 * * ] { } (2010) 036, [ [ 0911.4127]{}](http://arXiv.

]{}](htto://arXiv.org/abs/hep-th/9907219). Tomasitllo. D. L. Jafferis cbd A. Tokasielmo, “[A simole class of N=3 gauge/gravity vualw]{},” [[arXuv:0808.0864]{}](http://arXiv.org/abs/arXiv:0808.0864). A. Hanany wnd A. Zafdaroii, “[Tilings, Chern-Simons Tmzoriea and N2 Branes]{},” [*JHEP*]{} [**0810**]{} (2008) 111, [[ 0808.1244]{}](http://arFiv.org/abs/0808.1244). D. Martalui and J. Sparks, “[Notes on toric Sasaki-Eynstein sfven-manifolds wnd $SqS_4/CFF_3$]{},” [[arXiv:0808.0904]{}](http://arXiv.org/abs/arXiv:0808.0904). A. Hanang, D. Vegh, and A. Zaffaromi, “[Brane Tilings and M2 Brajes]{},” [ *JHEP*]{} [**0903**]{} (2009) 012, [[ 0809.1440]{}](http://arXig.org/abs/0809.1440). D. Marjslly and J. Sparks, “[AdS4/CFT3 duals from M2-brznes at hypersurface singularitkes aud their dedoemahhons]{},” [*JHEP*]{} [**12**]{} (2009) 017, [[0909.2036]{}](htti://arXiv.org/abs/0909.2036). S. Franco, H. R. Klebsnov, and D. Rodrlguez-Jomez, “[M2-branes on Orbifoldv of the Cone over Q\*\*1,1,1]{},” [*JHEP*]{} [**0908**]{} (2009) 033, [[0903.3231]{}](kttp://arXiv.org/abs/0903.3231). F. Benibi, C. Clovset, and W. Crdmohexi, “[Chirap fmavors and M2-branes at toric CY4 singulariuief]{},” [*JHEP*]{} [**1002**]{} (2010) 036, [[0911.4127]{}](http://zrXiv.

]{}](http://arXiv.org/abs/hep-th/9907219). Tomasiello. D. L. Jafferis and A. simple of N=3 duals]{},” [[arXiv:0808.0864]{}](http://arXiv.org/abs/arXiv:0808.0864). A. Chern-Simons and M2 Branes]{},” [**0810**]{} (2008) 111, 0808.1244]{}](http://arXiv.org/abs/0808.1244). D. Martelli and J. Sparks, on toric Sasaki-Einstein seven-manifolds and $AdS_4/CFT_3$]{},” [[arXiv:0808.0904]{}](http://arXiv.org/abs/arXiv:0808.0904). A. Hanany, D. Vegh, and A. “[Brane Tilings and M2 Branes]{},” [ *JHEP*]{} [**0903**]{} (2009) 012, [[ 0809.1440]{}](http://arXiv.org/abs/0809.1440). D. and Sparks, duals M2-branes at hypersurface singularities and their deformations]{},” [*JHEP*]{} [**12**]{} (2009) 017, [[0909.2036]{}](http://arXiv.org/abs/0909.2036). S. Franco, I. R. and D. Rodriguez-Gomez, “[M2-branes on Orbifolds of the over Q\*\*1,1,1]{},” [*JHEP*]{} [**0908**]{} 033, [[0903.3231]{}](http://arXiv.org/abs/0903.3231). F. Benini, C. and Cremonesi, “[Chiral and at CY4 singularities]{},” [*JHEP*]{} (2010) 036, [[0911.4127]{}](http://arXiv.

]{}](http://arXiv.org/abs/hep-th/9907219). TomasIello. D. L. JafFeris And a. ToMaSielLo, “[A sImple class of N=3 gAUge/gRavity duals]{},” [[arXiv:0808.0864]{}](http://arxiv.orG/aBS/arXIV:0808.0864). A. hananY and A. ZaFFaRONi, “[TIlInGs, CHeRN-SImons theOries anD M2 Branes]{},” [*JHeP*]{} [**0810**]{} (2008) 111, [[ 0808.1244]{}](hTtP://arXiv.org/abs/0808.1244). d. maRtelli and J. spaRks, “[Notes on toRic sasaki-eiNstEIn sevEn-mAnifoLds and $aDS_4/CFT_3$]{},” [[aRXiv:0808.0904]{}](http://aRXIV.org/abS/ArXiv:0808.0904). A. HANAnY, D. VeGh, and A. Zaffaroni, “[BrANe tIlings and M2 BranEs]{},” [ *JHEP*]{} [**0903**]{} (2009) 012, [[ 0809.1440]{}](HtTP://aRxIv.oRg/aBs/0809.1440). D. MartellI aNd J. SpARks, “[AdS4/Cft3 dUALS frOM M2-branes at hypErsurface siNGulAritieS aNd tHEir defOrmatIoNS]{},” [*JHeP*]{} [**12**]{} (2009) 017, [[0909.2036]{}](http://arXiv.Org/aBs/0909.2036). S. Franco, i. R. KlebANov, and D. rOdrigueZ-Gomez, “[m2-brAneS on ORBiFoLds Of THe CONe OveR q\*\*1,1,1]{},” [*JHeP*]{} [**0908**]{} (2009) 033, [[0903.3231]{}](http://arxiV.oRg/abs/0903.3231). f. BenINI, c. closSet, And S. cremoNesi, “[Chiral flaVorS and m2-BraNes at Toric cY4 siNgUlariTies]{},” [*JHeP*]{} [**1002**]{} (2010) 036, [[0911.4127]{}](httP://aRXiv.

]{}](http://arXiv.org/abs/ hep-th/990 7219) . T oma si ello . D.  L. Jafferis a n d A.  Tomasiello, “[A simpl e cla ss of N = 3gauge /gravit y d u a ls] {} ,” [[ ar X iv :0808 .08 64]{}]( http://arX iv. or g/abs/arXiv: 0 80 8.0864). A . H anany and A.  Za ffaron i, “[ T iling s,Chern -Simon s Theor ies and M 2B ranes] { },” [*J H E P* ]{}[**0810**]{} (200 8 )1 11, [[ 0808.12 44]{}] (h t tp : / /ar Xiv .org/abs/0 80 8.124 4 ). D.M ar t e l lia nd J. Sparks, “[Notes on tor ic Sas ak i-E i nstein seve n- m ani folds and $ AdS_ 4/CFT_3$] {},” [ [ arXiv:0 8 08.0904 ]{}](h ttp :// arXi v .o rg /ab s/ a rXi v :0 808 . 090 4). A. H an an y, D.  Veg h , a nd A . Z affa roni, “[Brane Tili ngs and M2Brane s]{}, ” [*J HEP*] {} [** 0903* *] {} (2009) 012,[[ 0 809.1440] {}] (h ttp :/ /arXi v .org/a bs/ 080 9.1440) . D. M a rte ll i a nd J. Sparks, “[AdS4 /C F T 3duals fr om M2- b ra ne s at hype rs urf aces i ngula riti e sand thei r defo r ma ti ons]{}, ”[*JHEP *] {}[** 12**] { } (2 009) 0 17, [[09 09.20 3 6]{}](http://a r Xiv.org/abs/0 9 09 . 2 03 6 ). S . F ranco, I. R . Kl e bano v, a n dD.R odrig uez-G om e z, “[M2-branes on Orbi fo lds of theCone over Q\* \*1,1,1]{} , ” [*JHEP*] {} [ * *0 9 08**]{} (2009) 033, [[0903.32 3 1]{}](ht tp:// arXiv.or g/abs/090 3 . 3231). F . B eni ni, C. C lo sset, and S.C r emon es i, “[Ch ira l flavo rsand M2 -br an es at tor ic CY4 s in gu la ri tie s]{}, ” [*JHEP* ]{ } [ ** 100 2**]{ } (2010 ) 036 , [[ 09 11 . 412 7]{}](h t tp : / /arX iv .

]{}](http://arXiv.org/abs/hep-th/9907219). Tomasiello._D. L. Jafferis_and A. Tomasiello, “[A simple_class of_N=3_gauge/gravity duals]{},”_[[arXiv:0808.0864]{}](http://arXiv.org/abs/arXiv:0808.0864)._A. Hanany and A. Zaffaroni,_“[Tilings, Chern-Simons Theories_and M2 Branes]{},” [*JHEP*]{}_[**0810**]{} (2008) 111,_[[_0808.1244]{}](http://arXiv.org/abs/0808.1244). D. Martelli and J. Sparks, “[Notes on toric Sasaki-Einstein seven-manifolds and $AdS_4/CFT_3$]{},” [[arXiv:0808.0904]{}](http://arXiv.org/abs/arXiv:0808.0904). A. Hanany, D. Vegh, and_A. Zaffaroni,_“[Brane Tilings_and_M2_Branes]{},” [ *JHEP*]{} [**0903**]{} (2009)_012, [[ 0809.1440]{}](http://arXiv.org/abs/0809.1440). D. Martelli and J. Sparks,_“[AdS4/CFT3 duals_from M2-branes at hypersurface singularities and their deformations]{},”_[*JHEP*]{}_[**12**]{} (2009) 017,_[[0909.2036]{}](http://arXiv.org/abs/0909.2036). S. Franco, I. R. Klebanov, and D. Rodriguez-Gomez, “[M2-branes on Orbifolds_of the Cone over Q\*\*1,1,1]{},” [*JHEP*]{}_[**0908**]{} (2009) 033,_[[0903.3231]{}](http://arXiv.org/abs/0903.3231)._F. Benini,_C. Closset, and S. Cremonesi, “[Chiral_flavors and M2-branes at toric CY4_singularities]{},” [*JHEP*]{} [**1002**]{} (2010) 036, [[0911.4127]{}](http://arXiv.

}$$ We say $(Y,Z,U)\in \mathbb D^{2,\kappa}_H \times \mathbb H^{2,\kappa}_H \times \mathbb J^{2,\kappa}_H$ is a solution to the $2$BSDEJ if - $Y_T=\xi$, $\mathcal{P}^\kappa_H$-q.s. - For all $\mathbb P \in \mathcal{P}^\kappa_H$ and $0\leq t\leq T$, the process $K^{\mathbb P}$ defined below is predictable and has non-decreasing paths $\mathbb P-a.s.$ $$K_t^{\mathbb P}:=Y_0-Y_t - \int_0^t\widehat{F}^{\mathbb P}_s(Y_s,Z_s,U_s)ds+\int_0^tZ_sdB^{\mathbb P,c}_s + \int_0^t \int_{E} U_s(x) {\mu}^{\mathbb P}_B(dx,ds). \label{2bsdej_K}$$ - The family $\left\{K^{\mathbb P}, \mathbb P \in \mathcal P_H^\kappa\right\}$ satisfies the minimum condition $$K_t^{\mathbb P}=\underset{ \mathbb{P}^{'} \in \mathcal{P}_H^\kappa(t^+,\mathbb{P}) }{ \einf^{\mathbb P} }\mathbb{E}_t^{\mathbb P^{'}}\left[K_T^{\mathbb{P}^{'}}\right], \text{ } 0\leq t\leq T, \text{ } \mathbb P-a.s., \text{ } \forall \mathbb P \in \mathcal P_H^\kappa. \label{2bsdej.minK}$$ If the family $\left\{K^{\mathbb P}, \mathbb P \in \mathcal P_H^\kappa\right\}$ can be aggregated into a universal process $K$, we call $(Y,Z,U,K)$ a solution of the $2$BSDEJ. Following [@stz], in addition to Assumption \[assump.href\], we will always assume \[assump.h2ref\] [(i)]{} $\mathcal P_H^\kappa$ is not empty. [(

} $ $ We say $ (Y, Z, U)\in \mathbb D^{2,\kappa}_H \times \mathbb H^{2,\kappa}_H \times \mathbb J^{2,\kappa}_H$ is a solution to the $ 2$BSDEJ if - $ Y_T=\xi$, $ \mathcal{P}^\kappa_H$-q.s. - For all $ \mathbb P \in \mathcal{P}^\kappa_H$ and $ 0\leq t\leq T$, the process $ K^{\mathbb P}$ defined downstairs is predictable and have non - decreasing paths $ \mathbb P - a.s.$ $ $ K_t^{\mathbb P}:=Y_0 - Y_t - \int_0^t\widehat{F}^{\mathbb P}_s(Y_s, Z_s, U_s)ds+\int_0^tZ_sdB^{\mathbb P, c}_s + \int_0^t \int_{E } U_s(x) { \mu}^{\mathbb P}_B(dx, ds). \label{2bsdej_K}$$ - The kin $ \left\{K^{\mathbb P }, \mathbb P \in \mathcal P_H^\kappa\right\}$ satisfies the minimal condition $ $ K_t^{\mathbb P}=\underset { \mathbb{P}^ {' } \in \mathcal{P}_H^\kappa(t^+,\mathbb{P }) } { \einf^{\mathbb P } } \mathbb{E}_t^{\mathbb P^{'}}\left[K_T^{\mathbb{P}^{'}}\right ], \text { } 0\leq t\leq T, \text { } \mathbb P - a.s. , \text { } \forall \mathbb P \in \mathcal P_H^\kappa. \label{2bsdej.minK}$$ If the family $ \left\{K^{\mathbb P }, \mathbb P \in \mathcal P_H^\kappa\right\}$ can be aggregate into a universal process $ K$, we address $ (Y, Z, U, K)$ a solution of the $ 2$BSDEJ. Following [ @stz ], in summation to Assumption \[assump.href\ ], we will constantly assume \[assump.h2ref\ ] [ (i) ] { } $ \mathcal P_H^\kappa$ is not empty. [ (

}$$ We day $(Y,Z,U)\in \mathbb D^{2,\kappa}_H \times \mathbb H^{2,\kappa}_H \vimes \mzthbb J^{2,\kxppa}_H$ is a solution to the $2$BDDWJ if - $Y_T=\xi$, $\mathcal{P}^\kappa_H$-d.s. - For wll $\mathvb P \un \mathcal{'}^\iappa_H$ and $0\les t\lex T$, the process $K^{\mathbb P}$ defined below ir 'redictable and has non-decreasing paehs $\matnbh P-a.s.$ $$K_t^{\mathbb P}:=Y_0-Y_u - \ynt_0^t\spdthat{F}^{\mathbb P}_s(Y_s,Z_s,U_s)ds+\int_0^tZ_sdB^{\matgbb P,c}_s + \int_0^t \int_{E} U_x(x) {\mu}^{\mathbb P}_B(dx,ds). \labep{2bsdfj_K}$$ - The family $\lfft\{K^{\mathbb K}, \maeybb P \in \matfcal P_H^\kapka\xight\}$ satisries the minimum condition $$K_t^{\maghbb '}=\underset{ \mqtybb{O}^{'} \in \mathcal{'}_H^\kappw(t^+,\mathbb{P}) }{ \elmf^{\mathtb P} }\mayhbb{E}_t^{\mathbb P^{'}}\kefv[K_T^{\mqthbb{P}^{'}}\right], \text{ } 0\leq t\leq T, \text{ } \matrbb P-a.s., \taxc{ } \forall \mathbb P \in \mqthcan P_H^\nappx. \lagek{2badej.mijK}$$ Ih the familg $\left\{K^{\mathvb P}, \mathbb P \in \mauhcwo P_H^\kappa\righf\}$ can fe aggregated into a universal process $K$, fe dall $(Y,Z,U,K)$ a solution of the $2$BSDEJ. Following [@sjz], in addieion to Assumption \[assump.href\], we will always assuke \[assnmo.h2rtf\] [(l)]{} $\oqtjcal P_H^\kappa$ is not empty. [(

}$$ We say $(Y,Z,U)\in \mathbb D^{2,\kappa}_H \times \times J^{2,\kappa}_H$ is solution to the - all $\mathbb P \mathcal{P}^\kappa_H$ and $0\leq T$, the process $K^{\mathbb P}$ defined is predictable and has non-decreasing paths $\mathbb P-a.s.$ $$K_t^{\mathbb P}:=Y_0-Y_t - \int_0^t\widehat{F}^{\mathbb P}_s(Y_s,Z_s,U_s)ds+\int_0^tZ_sdB^{\mathbb + \int_0^t \int_{E} U_s(x) {\mu}^{\mathbb P}_B(dx,ds). \label{2bsdej_K}$$ - The family $\left\{K^{\mathbb P}, \mathbb \in P_H^\kappa\right\}$ the condition $$K_t^{\mathbb P}=\underset{ \mathbb{P}^{'} \in \mathcal{P}_H^\kappa(t^+,\mathbb{P}) }{ \einf^{\mathbb P} }\mathbb{E}_t^{\mathbb P^{'}}\left[K_T^{\mathbb{P}^{'}}\right], \text{ } 0\leq t\leq T, } \mathbb P-a.s., \text{ } \forall \mathbb P \mathcal P_H^\kappa. \label{2bsdej.minK}$$ If family $\left\{K^{\mathbb P}, \mathbb P \mathcal can be into universal $K$, we call a solution of the $2$BSDEJ. Following [@stz], in addition to Assumption \[assump.href\], we will always assume \[assump.h2ref\] $\mathcal P_H^\kappa$ empty. [(

}$$ We say $(Y,Z,U)\in \mathbb D^{2,\kappa}_H \tiMes \mathbb H^{2,\Kappa}_h \tiMes \MaThbb j^{2,\kapPa}_H$ is a solution TO the $2$bSDEJ if - $Y_T=\xi$, $\mathcal{P}^\kapPa_H$-q.s. - foR All $\mAThBb P \in \Mathcal{p}^\KaPPA_H$ aNd $0\LeQ t\lEq t$, ThE procEss $k^{\mathbb p}$ defined beLow Is Predictable aND hAs non-decreAsiNg paths $\mathbB P-a.S.$ $$K_t^{\matHbB P}:=Y_0-y_T - \int_0^t\WidEhat{F}^{\Mathbb p}_S(Y_s,Z_s,U_S)ds+\int_0^tZ_sDB^{\MAthbb P,C}_S + \int_0^t \inT_{e} u_s(X) {\mu}^{\mAthbb P}_B(dx,ds). \label{2bSDeJ_k}$$ - The family $\left\{k^{\mathbB P}, \MAtHBB P \iN \maThcal P_H^\kapPa\Right\}$ SAtisfieS ThE MINimUM condition $$K_t^{\mAthbb P}=\underSEt{ \mAthbb{P}^{'} \In \MatHCal{P}_H^\kAppa(t^+,\MaTHbb{p}) }{ \einf^{\mathbb p} }\matHbb{E}_t^{\mathBb P^{'}}\lefT[k_T^{\mathbB{p}^{'}}\right], \tExt{ } 0\leq T\leQ T, \tExt{ } \mAThBb p-a.s., \TeXT{ } \foRAlL \maTHbb p \in \mathcAl p_H^\Kappa. \LabeL{2BSDEj.miNK}$$ IF the FamilY $\left\{K^{\mathbb P}, \MatHbb P \IN \maThcal p_H^\kapPa\riGhT\}$ can bE aggreGated InTo a universal proCess $k$, we call $(Y,Z,u,K)$ a SoLutIoN of thE $2$bSDEJ. FOllOwiNg [@stz], in AdditioN To ASsUMPTiOn \[assump.href\], we will AlWAYs Assume \[asSump.h2rEF\] [(i)]{} $\MaTHcal P_H^\kaPpA$ is Not eMPTy. [(

}$$ We say $(Y,Z,U)\in \m athbb D^{2 ,\kap pa} _H\t imes \ma thbb H^{2,\kap p a}_H \times \mathbb J^{2,\ kappa }_ H $ is asolut ion tot he $ 2$B SD EJ if - $Y_T =\x i$, $\m athcal{P}^ \ka pp a_H$-q.s. - For all $\ mat hbb P \in \m ath cal{P} ^\ kap p a_H$and $0\l eq t\l e q T$,the proce ss $K^{\m a thbb P} $ de fine d below is predic t ab l e and has non- decrea si n gp a ths $\ mathbb P-a .s .$ $$ K _t^{\ma t hb b P }:= Y _0-Y_t - \int _0^t\wideha t {F} ^{\mat hb b P } _s(Y_s ,Z_s, U_ s )ds +\int_0^tZ_ sdB^ {\mathbbP,c}_s + \int_ 0 ^t \int _{E} U _s( x){\mu } ^{ \m ath bb P}_ B (d x,d s ). \lab el {2 bsdej _K}$ $ - T hefami ly $\ left\{K^{\mat hbb P}, \ma thbbP \in \ma th cal P _H^\ka ppa\r ig ht\}$ satisfies the minimumcon di tio n$$K_t ^ {\math bbP}= \unders et{ \ma t hbb {P } ^ { '} \in \mathcal{P}_H ^\ k a pp a(t^+,\m athbb{ P }) } { \einf^{ \m ath bb P } }\mat hbb{ E }_ t^{\math bb P^{ ' }} \l eft[K_T ^{ \mathb b{ P}^ {'} }\rig h t],\text{ } 0\leq t\le q T, \text{ }\ mathbb P-a.s. , \ t e xt { } \ for all \mathbb P \ i n \m athc a lP_H ^ \kapp a. \ la b el{2bsdej.minK}$$ If the f amily $\left\{K^{\ mathbb P}, \ m athbb P\in\ ma t hcal P_H^\kapp a\rig ht\}$ canb e aggreg atedinto a u niversalp r ocess $K $,wecal l $ ( Y ,Z ,U,K)$ a solu t i on o fthe $2$ BSD EJ. Fo llo win g [ @st z] , in addi tion toAs su mp ti on\[ass u mp.href\ ], we w ill alwa y s assu me \ [ass um p. h 2re f\] [( i )] { } $\m at hc al P _H^ \k appa$ isn otempty. [(

}$$ We say_$(Y,Z,U)\in \mathbb_D^{2,\kappa}_H \times \mathbb H^{2,\kappa}_H_\times \mathbb_J^{2,\kappa}_H$_is a_solution_to the $2$BSDEJ_if - _$Y_T=\xi$, $\mathcal{P}^\kappa_H$-q.s. - _For all $\mathbb_P_\in \mathcal{P}^\kappa_H$ and $0\leq t\leq T$, the process $K^{\mathbb P}$ defined below is predictable_and_has non-decreasing_paths_$\mathbb_P-a.s.$ $$K_t^{\mathbb P}:=Y_0-Y_t - \int_0^t\widehat{F}^{\mathbb_P}_s(Y_s,Z_s,U_s)ds+\int_0^tZ_sdB^{\mathbb P,c}_s + \int_0^t \int_{E}_U_s(x) {\mu}^{\mathbb_P}_B(dx,ds). \label{2bsdej_K}$$ - The_family_$\left\{K^{\mathbb P}, \mathbb_P \in \mathcal P_H^\kappa\right\}$ satisfies the minimum condition $$K_t^{\mathbb_P}=\underset{ \mathbb{P}^{'} \in \mathcal{P}_H^\kappa(t^+,\mathbb{P}) }{ \einf^{\mathbb_P} }\mathbb{E}_t^{\mathbb P^{'}}\left[K_T^{\mathbb{P}^{'}}\right],_\text{_}_0\leq t\leq T, \text{_ } \mathbb P-a.s., \text{ }_\forall \mathbb P \in \mathcal P_H^\kappa. _ \label{2bsdej.minK}$$ If the family $\left\{K^{\mathbb_P}, \mathbb P \in \mathcal P_H^\kappa\right\}$_can be aggregated into a_universal process_$K$, we call $(Y,Z,U,K)$ a_solution of the_$2$BSDEJ. Following [@stz],_in addition to_Assumption \[assump.href\], we will always assume \[assump.h2ref\] [(i)]{}_$\mathcal P_H^\kappa$ is_not empty. [(

optimal solution to the problem by using a gradient-based solution (specifically using the function $\mathtt{fmincon}$ of the optimization toolbox of MATLAB). Note that the direction of $\bar{\mathbf{w}}_{dn}$ only affects the power of desired signal in the numerator of $\mathtt{SINR}_{n,FD}$. Therefore, the direction of the optimal $\bar{\mathbf{w}}_{dn}$ can be obtained as $\frac{\bar{\mathbf{w}}_{dn}^\star}{\|\bar{\mathbf{w}}_{dn}^\star\|} = \frac{\mathbf{g}_{Pkn}}{\|\mathbf{g}_{Pkn}\|}$, with which the number of variables in problem (\[E:problem-wb\]) is reduced. Practical Issues ================ By now, we have introduced the concept of fICIC and optimized the associated precoders. In this section, we discuss some practical issues regarding the application of the fICIC. Channel Acquisition ------------------- To apply the fICIC, the PBS needs to have the channels from the MBS to both the PBS and PUEs, i.e., ${\mathbf{H}}_{MP}$ and ${\mathbf{h}}_{Mk}$, as well as the channel from the PBS to PUEs, $\mathbf{h}_{Pk}$, $\forall k$. First, the channel ${\mathbf{H}}_{MP}$ can be directly estimated at the PBS by using the FD receive antennas to receive the downlink training signals sent from the MBS. Second, noting that in TDD systems channel reciprocity holds between the HD transceiver at the PBS and each PUE, the channel $\mathbf{h}_{Pk}$ can be estimated at the PBS by using the HD receive antennas to receive the uplink training signals sent by ${\text{PUE}}_k$. Finally, the channel ${\mathbf{h}}_{Mk}$ can be first estimated by ${\text{PUE}}_k$ and then fed back to the PBS, where digital or analog feedback schemes can be employed [@Marzetta2006]. We will evaluate the performance of the fICIC under imperfect channel estimation and feedback in next section. fICIC [v.s.]{} the HD Scheme ---------------------------- As we analyzed in Section \[S:asymptotic\], with the fICIC, the FD PBS can adaptively switch between FD mode and HD mode by optimizing the precoders for transmitting desired signals and forwarding

optimal solution to the problem by using a gradient - based solution (specifically use the affair $ \mathtt{fmincon}$ of the optimization toolbox of MATLAB). Note that the direction of $ \bar{\mathbf{w}}_{dn}$ only affect the power of desired signal in the numerator of $ \mathtt{SINR}_{n, FD}$. consequently, the direction of the optimal $ \bar{\mathbf{w}}_{dn}$ can be obtained as $ \frac{\bar{\mathbf{w}}_{dn}^\star}{\|\bar{\mathbf{w}}_{dn}^\star\| } = \frac{\mathbf{g}_{Pkn}}{\|\mathbf{g}_{Pkn}\|}$, with which the act of variables in problem (\[E: trouble - wb\ ]) is reduced. Practical Issues = = = = = = = = = = = = = = = = By now, we have introduced the concept of fICIC and optimized the associated precoders. In this incision, we discuss some practical issues regarding the application of the   fICIC. Channel Acquisition ------------------- To use the fICIC, the PBS needs to have the channels from the MBS to both the PBS and PUEs, i.e., $ { \mathbf{H}}_{MP}$ and $ { \mathbf{h}}_{Mk}$, equally well as the channel from the PBS to PUEs, $ \mathbf{h}_{Pk}$, $ \forall k$. First, the groove $ { \mathbf{H}}_{MP}$ can be directly estimate at the PBS by using the FD receive antennas to receive the downlink training signal sent from the MBS. Second, noting that in TDD systems channel reciprocity holds between the HD transceiver at the PBS and each PUE, the channel $ \mathbf{h}_{Pk}$ can be estimated at the PBS by using the HD receive antennas to receive the uplink training signals sent by $ { \text{PUE}}_k$. Finally, the channel $ { \mathbf{h}}_{Mk}$ can be first estimate by $ { \text{PUE}}_k$ and then fed back to the PBS, where digital or analog feedback system can be hire   [ @Marzetta2006 ]. We will evaluate the performance of the fICIC under fallible channel estimate and feedback in next section. fICIC [ v.s. ] { } the HD Scheme ---------------------------- As we analyze in Section \[S: asymptotic\ ], with the fICIC, the FD PBS can adaptively switch between FD mode and HD mode by optimizing the precoders for transmitting hope signals and forwarding

ophimal solution to the prublem by using c gradixnt-bases solutiun (specifically using the fuictiin $\mauktt{fmincon}$ of the opgimizatioj toolboz of NATLAB). Notx that tmz dirsgtion if $\bar{\mathbf{w}}_{dk}$ only affewts the power mf dzsired signal in the numerator of $\maehtt{SINT}_{n,VD}$. Therefore, tre dprqctikn of the optimal $\bar{\mathbf{w}}_{dn}$ can be obteined as $\frac{\bat{\mathbf{w}}_{dn}^\star}{\|\bar{\mathbf{w}}_{dn}^\dtar\|} = \frac{\mathbf{g}_{Pkn}}{\|\mahhbf{g}_{Pkn}\|}$, wijg wruch the numbdr of variables in proglem (\[E:problem-wb\]) is reduced. Practkcal Nssues ================ By noq, qe jdve introdured thv concept of nPCIC ang optimozed the assoclated prwcoders. In this sectimn, we discuss some practican nssues regarding the qpplicajion mf tfw fIZIC. Dhennsl Acqkismtion ------------------- To appmy the fICIX, the PBS needs to nadv the channela from tre MBS to both the PBS and PUEs, i.e., ${\mathtf{H}}_{JP}$ and ${\mathbf{h}}_{Mk}$, as welo as the channel from the PBS eo PUEs, $\mathbf{h}_{Pk}$, $\forall k$. First, the channel ${\mathtf{H}}_{MP}$ zan bc difwchly estimated at the PBS by using the FD receybe aktennas to receiye the downlink trsijimd signals senj from cge MBS. Second, noting that ig TDD systems shanmel reciprocity holds betwewn the HD trcnsxeiver at the PBS cnd each PUE, the chanmel $\mathbf{h}_{Pk}$ can be escimates at the PBD by usinf the HD receive xntvnnav to rectkve the uplink trwining sijnals sent bh ${\tect{PUE}}_k$. Finally, tje channel ${\mathbf{h}}_{Mk}$ can be fnrst astimated hy ${\text{PUE}}_k$ and then fed back to the PBS, wherg dhginal or anclog fcedback schemes can be employgd [@Marzettc2006]. We wkll evaluane the pecformance of the fICIC unggr imperfect rhannel eftimqtiob and fdddback in next section. fPCNC [v.s.]{} the HD Scheme ---------------------------- As we anelyzgd in Section \[S:asvoktitic\], with the gICKC, ehv FV PBS wan adaptiveny sditzn betdeen FD modc avd HC mode by optimizing the precoders for tramsiitting eesired fignals and fprwarding

optimal solution to the problem by using solution using the $\mathtt{fmincon}$ of the that direction of $\bar{\mathbf{w}}_{dn}$ affects the power desired signal in the numerator of Therefore, the direction of the optimal $\bar{\mathbf{w}}_{dn}$ can be obtained as $\frac{\bar{\mathbf{w}}_{dn}^\star}{\|\bar{\mathbf{w}}_{dn}^\star\|} = with which the number of variables in problem (\[E:problem-wb\]) is reduced. Practical Issues By we introduced concept of fICIC and optimized the associated precoders. In this section, we discuss some practical issues the application of the fICIC. Channel Acquisition ------------------- apply the fICIC, the needs to have the channels the to both PBS PUEs, ${\mathbf{H}}_{MP}$ and ${\mathbf{h}}_{Mk}$, well as the channel from the PBS to PUEs, $\mathbf{h}_{Pk}$, $\forall k$. First, the channel ${\mathbf{H}}_{MP}$ can directly estimated PBS by the receive to receive the signals sent from the MBS. Second, TDD systems channel reciprocity holds between the HD at the and each PUE, the channel $\mathbf{h}_{Pk}$ be estimated at the PBS by using the receive antennas to receive the uplink training signals sent by ${\text{PUE}}_k$. Finally, the channel ${\mathbf{h}}_{Mk}$ first estimated by ${\text{PUE}}_k$ then fed back the where or feedback schemes be employed [@Marzetta2006]. We will evaluate the performance of the fICIC imperfect channel estimation and feedback in next section. fICIC [v.s.]{} Scheme As we analyzed Section \[S:asymptotic\], with the the PBS can adaptively switch mode HD the for desired signals and forwarding

optimal solution to the problEm by using a GradiEnt-BasEd SoluTion (Specifically usINg thE function $\mathtt{fmincon}$ Of the OpTImizATiOn tooLbox of MatLab). notE tHaT thE dIReCtion Of $\bAr{\mathbF{w}}_{dn}$ only afFecTs The power of deSIrEd signal in The Numerator of $\mAthTt{SINR}_{N,Fd}$. ThEReforE, thE direCtion oF The optImal $\bar{\maThBF{w}}_{dn}$ caN Be obtaiNED aS $\fraC{\bar{\mathbf{w}}_{dn}^\star}{\|\BAr{\MAthbf{w}}_{dn}^\star\|} = \frAc{\mathBf{G}_{pkN}}{\|\MAthBf{g}_{pkn}\|}$, with whiCh The nuMBer of vaRIaBLES in PRoblem (\[E:probleM-wb\]) is reduceD. praCtical isSueS ================ by now, wE have InTRodUced the concEpt oF fICIC and OptimiZEd the asSOciated PrecodErs. in tHis sECtIoN, we DiSCusS SoMe pRActIcal issuEs ReGardiNg thE APPLicaTioN of tHe fICiC. Channel AcquIsiTion ------------------- tO apPly thE fICIc, the pBs needS to havE the cHaNnels from the MBS To boTh the PBS aNd PuES, i.e., ${\MaThbf{H}}_{mp}$ and ${\maThbF{h}}_{MK}$, as well As the chANneL fROM ThE PBS to PUEs, $\mathbf{h}_{PK}$, $\fORAlL k$. First, tHe chanNEl ${\MaTHbf{H}}_{MP}$ caN bE diRectLY EstimAted AT tHe PBS by uSing thE fD ReCeive anTeNnas to ReCeiVe tHe dowNLink TrainiNg signalS sent FRom the MBS. SeconD, Noting that in Tdd sYSTeMS chaNneL reciprocitY holDS betWeen THe hD tRAnsceIver aT tHE Pbs and each PUE, the channEl $\Mathbf{H}_{Pk}$ caN be estimated aT the PBS by uSING the HD reCeivE AnTEnnas to receive The upLink trainiNG signals Sent bY ${\text{PUE}}_K$. Finally, tHE Channel ${\mAthBf{h}}_{mk}$ cAn bE FIrSt estimated by ${\TEXt{PUe}}_k$ And then Fed Back to tHe PbS, wHerE diGiTal or analOg feedbaCk ScHeMeS caN be emPLoyed [@MarZeTta2006]. we WilL evalUAte the PerfoRmanCe Of THe fiCIC undER iMPErfeCt ChAnneL esTiMatioN and FEedBack in nExt sectioN. fIciC [v.s.]{} ThE Hd Scheme ---------------------------- as we analyzed iN SEction \[S:asyMpTotIc\], with THE fICIC, thE FD PBS can adaptively switCH betweeN FD Mode aNd HD Mode by optImiZing thE prECoders For traNsmitTiNg dESIred sIGNaLs aNd Forwarding

optimal solution to the p roblem byusing agra di ent- base d solution (sp e cifi cally using the functi on $\ ma t htt{ f mi ncon} $ of th e o p t imi za ti onto o lb ox of MA TLAB).Note thatthe d irection of$ \b ar{\mathbf {w} }_{dn}$ only af fectsth e p o wer o f d esire d sign a l in t he numera to r of $\ m athtt{S I N R} _{n, FD}$. Therefore,t he direction of t he opt im a l$ \ bar {\m athbf{w}}_ {d n}$ c a n be ob t ai n e d as $\frac{\bar{\ mathbf{w}}_ { dn} ^\star }{ \|\ b ar{\ma thbf{ w} } _{d n}^\star\|} = \ frac{\mat hbf{g} _ {Pkn}}{ \ |\mathb f{g}_{ Pkn }\| }$,w it hwhi ch the nu mbe r of variabl es i n pro blem ( \ [ E:pr obl em-w b\])is reduced. Pra ctic a l I ssues ==== ==== == ===== = Bynow,we have introduce d th e concept of f ICI Cand o p timize d t heassocia ted pre c ode rs . I nthis section, we d is c u ss some pr actica l i ss u es regar di ngthea p plica tion of the fIC IC. C h an ne l Acqui si tion - -- --- --- ----- - ---- To a pply the fICI C , the PBS need s to have thec ha n n el s fro m t he MBS to b otht he P BS a n dPUE s , i.e ., ${ \m a th b f{H}}_{MP}$ and ${\ ma thbf{h }}_{M k}$, as wellas the cha n n e l from t he P B St o PUEs, $\math bf{h} _{Pk}$, $\ f orall k$ . Fir st, thechannel $ { \ mathbf{H }}_ {MP }$can b edirectly esti m a tedat the PB S b y using th e F D r ece iv e antenna s to rec ei ve t he do wnlin k trainin gsig na lssentf rom th e MBS . Se co nd , no ting th a ti n TDD s ys tems ch an nel r ecip r oci ty hold s between th e HDtr an sceiver at the PBS a nd each PUE, t hechanne l $\mathbf {h}_{Pk}$ can be estima t ed at t hePBS b y us ing the H D r eceive an t ennasto rec eiveth e u p l ink t r a in ing s ignals sen t by${\te xt {PUE }}_k$.Finally, the chann e l $ {\mathbf{h}}_ {Mk }$ c a n b e f i rs t es ti m ate d by ${\text{PUE} }_k$ and t he n f ed back to the P BS, whe re digi tal o r analog feedback schemesca n be e mpl oyed [@Mar zetta200 6]. We wi l l eva l ua te th e p erform an ceof th e fICI C un der i mperfe ct chann el es ti mation a nd feedback in next sec tion. fICI C [ v.s.]{} t heH D S cheme --- ---- ---------- --- --- ----- A s we a naly z ed in Secti on \ [ S:asympto t ic \], w it h the fICIC , t heFD PB S c a n adap tive ly switch between FD mode and HD mod e byopt i mizi ng the precoders fo rt r ansmitti ng desired si gnals an df orwar ding

optimal_solution to_the problem by using_a gradient-based_solution_(specifically using_the_function $\mathtt{fmincon}$ of_the optimization toolbox_of MATLAB). Note that_the direction of_$\bar{\mathbf{w}}_{dn}$_only affects the power of desired signal in the numerator of $\mathtt{SINR}_{n,FD}$. Therefore, the_direction_of the_optimal_$\bar{\mathbf{w}}_{dn}$_can be obtained as $\frac{\bar{\mathbf{w}}_{dn}^\star}{\|\bar{\mathbf{w}}_{dn}^\star\|}_= \frac{\mathbf{g}_{Pkn}}{\|\mathbf{g}_{Pkn}\|}$, with which the_number of_variables in problem (\[E:problem-wb\]) is reduced. Practical Issues ================ By now,_we_have introduced the_concept of fICIC and optimized the associated precoders. In_this section, we discuss some practical_issues regarding the_application_of_the fICIC. Channel Acquisition ------------------- To apply the_fICIC, the PBS needs to have_the channels from the MBS to_both the PBS and PUEs, i.e., ${\mathbf{H}}_{MP}$_and ${\mathbf{h}}_{Mk}$, as well as the_channel from the PBS to_PUEs, $\mathbf{h}_{Pk}$,_$\forall k$. First, the channel_${\mathbf{H}}_{MP}$ can be_directly estimated_at the PBS_by using the FD receive antennas_to receive the_downlink training signals sent from the_MBS._Second, noting that_in_TDD_systems channel_reciprocity holds between_the_HD transceiver_at_the PBS and each PUE, the_channel_$\mathbf{h}_{Pk}$ can be estimated at the PBS_by using the HD_receive_antennas to receive the_uplink training signals sent by_${\text{PUE}}_k$. Finally, the channel ${\mathbf{h}}_{Mk}$ can_be first_estimated by_${\text{PUE}}_k$ and then fed back to the PBS, where digital or_analog feedback schemes can be employed [@Marzetta2006]._We will evaluate the_performance of_the_fICIC under imperfect_channel_estimation and_feedback in next section. fICIC [v.s.]{} the HD_Scheme ---------------------------- As we_analyzed in Section \[S:asymptotic\], with the_fICIC, the FD PBS_can_adaptively switch between FD mode and_HD mode by optimizing the precoders_for transmitting desired signals and_forwarding

G\operatorname{Res}_G^U(B),B)$ because ${\mathrm{KK}}$ is stable. One verifies easily that the morphisms $\pi_B$ and $\iota_A$ are units of adjunction, so that we get . After these purely formal manipulations of functors, we now come to a much deeper assertion, which is due to Gennadi Kasparov. \[pro:ind\_res\] Let $G$ be almost connected and let $H\subseteq G$ be a maximal compact subgroup. If one of $X$, $A$ and $B$ is a proper $G$[-]{}space or a proper $G$-$C^*$[-]{}algebra, then $$\label{eq:res_mc} \operatorname{Res}_G^H\colon {\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(A,B) \to {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_G^H A,\operatorname{Res}_G^H B)$$ is an isomorphism. \[lem:ind\_res\] Let $H\subseteq G$ be a large compact subgroup and let $U{\mathrel{:=}}U_H$. There is a natural isomorphism $$\label{eq:ind_res} {\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(\operatorname{Ind}_H^G A,B) \cong {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_U^H \operatorname{Ind}_H^U A,\operatorname{Res}_G^H B).$$ Define $J_H^G(A){\mathrel{:=}}\operatorname{Ind}_H^G (C_0((U/H)^7)\otimes A)$. Then there is a natural isomorphism $${\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(J_H^G A,B)\cong {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(A,\operatorname{Res}_G^H B)$$ if $A$ belongs to the essential range of the functor $\operatorname{Res}_U^H$. Furthermore, the functors $$\operatorname{Res}_U^H\colon {\mathrm{KK}}^{U{\mathbin{\ltimes}}X}\to{\mathrm{KK}}^{H{\mathbin{\ltimes

G\operatorname{Res}_G^U(B),B)$ because $ { \mathrm{KK}}$ is stable. One verifies easily that the morphisms $ \pi_B$ and $ \iota_A$ are units of adjunction, so that we draw  . After these strictly formal manipulations of functors, we nowadays issue forth to a much deeper affirmation, which is due to Gennadi Kasparov. \[pro: ind\_res\ ] Let   $ G$ be almost get in touch and let $ H\subseteq G$ be a maximal compact subgroup. If one of $ X$, $ A$ and   $ B$ is a proper $ G$[-]{}space or a proper $ G$-$C^*$[-]{}algebra, then $ $ \label{eq: res_mc } \operatorname{Res}_G^H\colon { \mathrm{KK}}^{G{\mathbin{\ltimes}}X}(A, B) \to { \mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_G^H A,\operatorname{Res}_G^H B)$$ is an isomorphism. \[lem: ind\_res\ ] lease $ H\subseteq G$ be a large compact subgroup and lease $ U{\mathrel{:=}}U_H$. There is a natural isomorphism $ $ \label{eq: ind_res } { \mathrm{KK}}^{G{\mathbin{\ltimes}}X}(\operatorname{Ind}_H^G A, B) \cong { \mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_U^H \operatorname{Ind}_H^U A,\operatorname{Res}_G^H B).$$ Define $ J_H^G(A){\mathrel{:=}}\operatorname{Ind}_H^G (C_0((U / H)^7)\otimes A)$. Then there is a natural isomorphism $ $ { \mathrm{KK}}^{G{\mathbin{\ltimes}}X}(J_H^G A, B)\cong { \mathrm{KK}}^{H{\mathbin{\ltimes}}X}(A,\operatorname{Res}_G^H B)$$ if   $ A$ belong to the essential range of the functor $ \operatorname{Res}_U^H$. Furthermore, the functors $ $ \operatorname{Res}_U^H\colon { \mathrm{KK}}^{U{\mathbin{\ltimes}}X}\to{\mathrm{KK}}^{H{\mathbin{\ltimes

G\opfratorname{Res}_G^U(B),B)$ becaust ${\mathrm{KK}}$ is stayoe. One verifjes easiuy that the morphisms $\pi_B$ anv $\iora_A$ aee units of adjunction, so that ae get . Afrer uhese purely formem manipmjatikks of hunctors, we now come to a much deeper avsdrcion, which is due to Gennadi Kasparod. \[pro:ind\_ted\] Let $G$ be almoft cpgnecfvd and let $H\subseteq G$ be a maxjmal cokpact subgroul. If one of $X$, $A$ and $B$ is a oropfr $G$[-]{}space or a prooer $G$-$C^*$[-]{}algebtz, trwn $$\label{eq:rer_mc} \opegctorname{Res}_F^H\colon {\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(X,B) \to {\mathrm{JK}}^{Y{\mahvbin{\ltimes}}X}(\iperanorname{Res}_G^H A,\operatosname{Rex}_G^H B)$$ is an ispmocphiwm. \[lem:ind\_res\] Let $H\subsxteq G$ be a large coipact subcrkup and let $U{\matheeo{:=}}U_H$. Tvere is x nagurzl iaomorpjisj $$\label{eq:ihd_res} {\marhrm{KK}}^{G{\mathbin{\ltimex}}X}(\jirratorname{Ins}_H^G A,B) \cong {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorndme{Des}_U^H \operatorname{Ind}_H^U A,\operatorname{Res}_G^H B).$$ Define $J_R^G(A){\mathrel{:=}}\operatorname{Ind}_H^G (C_0((U/H)^7)\otimes A)$. Then thera is e vatbvwu idomorphism $${\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(J_H^G A,B)\cong {\iztnrk{KK}}^{H{\mathbin{\ltiies}}X}(A,\operatprjakg{Res}_G^H B)$$ if $A$ bglongs ck fhe essential rangf of thg funcror $\operauornake{Res}_U^H$. Furthermore, the funxtors $$\operatjename{Res}_U^H\colon {\machrm{KK}}^{U{\mathbnn{\ltimgs}}X}\to{\msthrm{KK}}^{H{\mathbin{\ltimes

G\operatorname{Res}_G^U(B),B)$ because ${\mathrm{KK}}$ is stable. One verifies the $\pi_B$ and are units of . these purely formal of functors, we come to a much deeper assertion, is due to Gennadi Kasparov. \[pro:ind\_res\] Let $G$ be almost connected and let G$ be a maximal compact subgroup. If one of $X$, $A$ and $B$ a $G$[-]{}space a $G$-$C^*$[-]{}algebra, then $$\label{eq:res_mc} \operatorname{Res}_G^H\colon {\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(A,B) \to {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_G^H A,\operatorname{Res}_G^H B)$$ is an isomorphism. \[lem:ind\_res\] Let $H\subseteq G$ a large compact subgroup and let $U{\mathrel{:=}}U_H$. There a natural isomorphism $$\label{eq:ind_res} A,B) \cong {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_U^H \operatorname{Ind}_H^U A,\operatorname{Res}_G^H Define (C_0((U/H)^7)\otimes A)$. there a isomorphism $${\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(J_H^G A,B)\cong B)$$ if $A$ belongs to the essential range of the functor $\operatorname{Res}_U^H$. Furthermore, the functors $$\operatorname{Res}_U^H\colon {\mathrm{KK}}^{U{\mathbin{\ltimes}}X}\to{\mathrm{KK}}^{H{\mathbin{\ltimes

G\operatorname{Res}_G^U(B),B)$ becauSe ${\mathrm{KK}}$ Is staBle. one VeRifiEs eaSily that the morPHismS $\pi_B$ and $\iota_A$ are units of AdjunCtIOn, so THaT we geT . After tHEsE PUreLy FoRmaL mANiPulatIonS of funcTors, we now cOme To A much deeper aSSeRtion, which Is dUe to Gennadi KAspArov. \[prO:iNd\_rES\] Let $G$ Be aLmost ConnecTEd and lEt $H\subsetEq g$ Be a maxIMal compACT sUbgrOup. If one of $X$, $A$ and $B$ iS A pROper $G$[-]{}space or a pRoper $G$-$c^*$[-]{}aLGeBRA, thEn $$\lAbel{eq:res_mC} \oPeratORname{ReS}_g^H\COLOn {\mAThrm{KK}}^{G{\mathbiN{\ltimes}}X}(A,B) \tO {\MatHrm{KK}}^{H{\MaThbIN{\ltimeS}}X}(\opeRaTOrnAme{Res}_G^H A,\opEratOrname{Res}_g^H B)$$ is aN IsomorpHIsm. \[lem:iNd\_res\] LEt $H\SubSeteQ g$ bE a LarGe COmpACt SubGRouP and let $U{\MaThRel{:=}}U_H$. therE IS A NatuRal IsomOrphiSm $$\label{eq:ind_rEs} {\mAthrM{kK}}^{G{\MathbIn{\ltiMes}}X}(\OpEratoRname{INd}_H^G A,b) \cOng {\mathrm{KK}}^{H{\matHbin{\Ltimes}}X}(\opEraToRnaMe{res}_U^H \OPeratoRnaMe{INd}_H^U A,\opEratornAMe{REs}_g^h b).$$ deFine $J_H^G(A){\mathrel{:=}}\opeRaTORnAme{Ind}_H^G (c_0((U/H)^7)\otiMEs a)$. THEn there iS a NatUral ISOmorpHism $${\MAtHrm{KK}}^{G{\maThbin{\lTImEs}}x}(J_H^G A,B)\cOnG {\mathrM{Kk}}^{H{\mAthBin{\ltIMes}}X}(a,\operaTorname{REs}_G^H B)$$ IF $A$ belongs to the ESsential range OF tHE FuNCtor $\OpeRatorname{ReS}_U^H$. FURtheRmorE, ThE fuNCtors $$\OperaToRNaME{Res}_U^H\colon {\mathrm{KK}}^{u{\mAthbin{\LtimeS}}X}\to{\mathrm{KK}}^{H{\Mathbin{\ltiMES

G\operatorname{Res}_G^U(B) ,B)$ becau se ${ \ma thr m{ KK}} $ is stable. One v e rifi es easily that the mor phism s$ \pi_ B $and $ \iota_A $ a r e un it sofad j un ction , s o thatwe get . Aft er these purel y f ormal mani pul ations of fu nct ors, w enow cometoa muc h deep e r asse rtion, wh ic h is du e to Gen n a di Kas parov. \[pro:ind \ _r e s\] Let $G$ be almos tc on n e cte d a nd let $H\ su bsete q G$ bea m a x i mal compact subgr oup. If one of$X$, $ A$ an d  $B$ i s a p ro p er$G$[-]{}spa ce o r a prope r $G$- $ C^*$[-] { }algebr a, the n $ $\l abel { eq :r es_ mc } \ ope r ato rname{Re s} _G ^H\co lon{ \ m a thrm {KK }}^{ G{\ma thbin{\ltimes }}X }(A, B ) \t o {\m athr m{ KK}}^ {H{\ma thbin {\ ltimes}}X}(\ope rato rname{Res }_G ^H A, \o perat o rname{ Res }_G ^H B)$$ is ani som or p h i sm . \[lem:ind\_res\ ]L e t$H\subse teq G$ be a large co mp act sub g r oup a nd l e t$U{\math rel{:= } }U _H $. Ther eis a n at ura l i somor p hism $$\la bel{eq:i nd_re s } {\mathrm { KK}}^{G{\math b in { \ lt i mes} }X} (\operatorn ame{ I nd}_ H^GA ,B ) \co ng {\ ma t hr m {KK}}^{H{\mathbin{\ lt imes}} X}(\o peratorname{R es}_U^H \o p e r atorname {Ind } _H ^ U A,\operatorn ame{R es}_G^H B) . $$ Defin e $J_ H^G(A){\ mathrel{: = } }\operat orn ame {In d}_ H ^ G(C_0((U/H)^7) \ o time sA)$. Th enthere i s a na tur alis omorphism $${\mat hr m{ KK }} ^{G {\mat h bin{\lti me s}} X} (J_ H^G A , B)\con g {\m athr m{ KK } }^{ H{\math b in { \ ltim es }} X}(A ,\o pe rator name { Res }_G^H B )$$ if $A $ b e long sto the es sential range o f the func to r $ \opera t o rname{Re s}_U^H$. Furthermore, t h e funct ors $$\o pera torname{R es} _U^H\c olo n {\mat hrm{KK }}^{U {\ mat h b in{\l t i me s}} X} \to{\mathr m { KK} }^{H{ \m athb in{\lti mes

G\operatorname{Res}_G^U(B),B)$ because_${\mathrm{KK}}$ is_stable. One verifies easily_that the_morphisms_$\pi_B$ and_$\iota_A$_are units of_adjunction, so that_we get . After these purely_formal manipulations of_functors,_we now come to a much deeper assertion, which is due to Gennadi Kasparov. \[pro:ind\_res\]_Let $G$_be almost_connected_and_let $H\subseteq G$ be a_maximal compact subgroup. If one_of $X$,_$A$ and $B$ is a proper $G$[-]{}space or a_proper_$G$-$C^*$[-]{}algebra, then $$\label{eq:res_mc} _ \operatorname{Res}_G^H\colon {\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(A,B) \to_{\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_G^H A,\operatorname{Res}_G^H B)$$ is an isomorphism. \[lem:ind\_res\]_Let $H\subseteq G$_be_a_large compact subgroup and_let $U{\mathrel{:=}}U_H$. There is a natural_isomorphism $$\label{eq:ind_res} {\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(\operatorname{Ind}_H^G_A,B) \cong {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(\operatorname{Res}_U^H \operatorname{Ind}_H^U_A,\operatorname{Res}_G^H B).$$ Define $J_H^G(A){\mathrel{:=}}\operatorname{Ind}_H^G (C_0((U/H)^7)\otimes A)$._Then there is a natural_isomorphism $${\mathrm{KK}}^{G{\mathbin{\ltimes}}X}(J_H^G_A,B)\cong {\mathrm{KK}}^{H{\mathbin{\ltimes}}X}(A,\operatorname{Res}_G^H B)$$ if $A$ belongs_to the essential_range of_the functor $\operatorname{Res}_U^H$._Furthermore, the functors $$\operatorname{Res}_U^H\colon {\mathrm{KK}}^{U{\mathbin{\ltimes}}X}\to{\mathrm{KK}}^{H{\mathbin{\ltimes

$y>0$, $${{\mathbb P}}^{y} \{M\geq x \} \leq \frac{(6+\varepsilon)}{\sigma^{2}x^{2}}$$ for all sufficiently large $x$. It is by no means evident, however, that the tail behavior should be the same as for branching Brownian motion with no killing. In fact we will prove that when the branching process is initiated by a single particle at a location $y>0$ near zero, the tail follows a power law with exponent $3$ rather than $2$. In particular, we will prove in Theorem \[thm.max.general\] that for each fixed $y>0$, $${{\mathbb P}}^y (M>x)\sim \frac{C_3y}{x^3} \quad \text{as} \;\; x \rightarrow \infty.$$ where $C_{3}>0$ is a constant depending on the offspring distribution but not on $x$ or $y$. On the other hand, we will show that for initial particle locations $y=sx$ whose distances from the killing zone are proportional to the target $x$, the exponent of the power law reverts to $2$; in particular, there exists a continuous function $C_{4} (s)$ of $s\in (0,1)$ such that $${{\mathbb P}}^{sx} (M>x)\sim \frac{C_4 (s)}{x^2} \quad \text{as} \;\; x \rightarrow \infty.$$ Furthermore, we will show that in the *Moranian* case, where the offspring law is double-or-nothing, the tail probability ${{\mathbb P}}^y (M>x)$ can be explicitly written as a Weierstrass $\mathcal{P}-$function. All of these results will be deduced from an analysis of a boundary value problem in the variable $y$ satisfied by the hitting probability ${{\mathbb P}}^y (M>x)$. Also of interest is the total number $N$ of particles killed at $0$. For supercritical branching Brownian motion with particle drift and killing at $0$, Maillard [@M13] and Berestycki *et al.* [@BBHM15] have, under various hypotheses concerning the drift and the reproduction mechanism, obtained sharp estimates for the tail of the distribution of $N$. For critical branching Brownian motion with killing, T. Y. Lee [@L90-1

$ y>0 $, $ $ { { \mathbb P}}^{y } \{M\geq x \ } \leq \frac{(6+\varepsilon)}{\sigma^{2}x^{2}}$$ for all sufficiently large $ x$. It is by no means evident, however, that the tail demeanor should be the like as for branch Brownian motion with no killing. In fact we will rise that when the branching summons is initiated by a single atom at a location $ y>0 $ near zero, the tail follows a ability law with exponent $ 3 $ quite than $ 2$. In particular, we will prove in Theorem \[thm.max.general\ ] that for each specify $ y>0 $, $ $ { { \mathbb P}}^y (M > x)\sim \frac{C_3y}{x^3 } \quad \text{as } \;\; x \rightarrow \infty.$$ where $ C_{3}>0 $ is a constant depending on the offspring distribution but not on $ x$ or $ y$. On the other hand, we will prove that for initial particle locations $ y = sx$ whose distances from the killing zone are proportional to the aim $ x$, the exponent of the power law reverts to $ 2 $; in especial, there exists a continuous function $ C_{4 } (s)$ of $ s\in (0,1)$ such that $ $ { { \mathbb P}}^{sx } (M > x)\sim \frac{C_4 (s)}{x^2 } \quad \text{as } \;\; x \rightarrow \infty.$$ Furthermore, we will show that in the * Moranian * case, where the offspring jurisprudence is double - or - nothing, the tail probability $ { { \mathbb P}}^y (M > x)$ can be explicitly written as a Weierstrass $ \mathcal{P}-$function. All of these results will be deduced from an analysis of a boundary value problem in the variable $ y$ satisfied by the hitting probability $ { { \mathbb P}}^y (M > x)$. Also of interest is the total number $ N$ of particles killed at $ 0$. For supercritical ramify Brownian motion with particle drift and kill at $ 0 $, Maillard [ @M13 ] and Berestycki * et al. * [ @BBHM15 ] have, under diverse hypotheses concerning the drift and the reproduction mechanism, prevail sharp estimates for the tail of the distribution of $ N$. For critical branching Brownian apparent motion with killing, T.   Y.   Lee [ @L90 - 1

$y>0$, $${{\lathbb P}}^{y} \{M\geq x \} \leq \frxc{(6+\varepsilon)}{\sigma^{2}x^{2}}$$ for all shfficiengly large $x$. It is by no means ecidenu, however, that the gail behanior shouod bt the same as for branchiky Broskian kition with no lilling. In fact we will [ruvz that when the branching process is initiayef by a single kartiblq at a location $y>0$ near zero, the tail fkllows e power law witn exponent $3$ rather than $2$. Ij pagticular, we will pgove in Theirem \[rhm.max.generau\] that for each fixed $i>0$, $${{\mathbb P}}^y (M>x)\sim \frac{C_3y}{x^3} \quad \tdxt{as} \;\; x \rigytqrrlf \infty.$$ whece $C_{3}>0$ if a constant dependitg on tne offspring dlstriuutiin but not on $x$ or $y$. Mn the other hand, re will svoc that for initial paetucle nocadionr $y=sb$ wgoxe distajcea from the killing zobe are proportional tj the target $x$, the evpjnent of the power law reverts to $2$; in pdrtjcular, there exists a cintinuous function $C_{4} (d)$ of $s\in (0,1)$ such that $${{\mathbb P}}^{sx} (M>x)\sim \frac{C_4 (s)}{x^2} \quad \text{as} \;\; x \figkbwfeoa \infty.$$ Furthermore, we will show that in the *Ikrsnpan* case, where thc offspring law is dlunje-or-nothing, tfe tail pdobability ${{\mathbb O}}^y (M>x)$ cag be wxplicitlr wrotten as a Weierstrass $\mathxal{P}-$function. Qll of these resulcs will be dzduced from an analysis of a boundcry vamue problem in the vzfiable $y$ satisfidd ny dhe hitting probability ${{\maehbb P}}^y (M>e)$. Also of intdresj is thq total nulber $K$ of particles killfd at $0$. Xor supercgitical branching Brownian motion with particke drpft and knlling at $0$, Maillard [@M13] and Berestyeki *et aj.* [@BBHO15] have, undvr variouv hypothesef concerning dje drift and the repwoduxtiob mechavksm, obtained snarp estimates for rhe tail of the dixtrkgution of $N$. For crutical branchinb Bfowgiwn mjdion with kinline, T. H. Kee [@L90-1

$y>0$, $${{\mathbb P}}^{y} \{M\geq x \} \leq all large $x$. is by no tail should be the as for branching motion with no killing. In fact will prove that when the branching process is initiated by a single particle a location $y>0$ near zero, the tail follows a power law with exponent rather $2$. particular, will prove in Theorem \[thm.max.general\] that for each fixed $y>0$, $${{\mathbb P}}^y (M>x)\sim \frac{C_3y}{x^3} \quad \text{as} x \rightarrow \infty.$$ where $C_{3}>0$ is a constant on the offspring distribution not on $x$ or $y$. the hand, we show for particle locations $y=sx$ distances from the killing zone are proportional to the target $x$, the exponent of the power law to $2$; there exists continuous $C_{4} of $s\in (0,1)$ $${{\mathbb P}}^{sx} (M>x)\sim \frac{C_4 (s)}{x^2} \quad \rightarrow \infty.$$ Furthermore, we will show that in *Moranian* case, the offspring law is double-or-nothing, the probability ${{\mathbb P}}^y (M>x)$ can be explicitly written a Weierstrass $\mathcal{P}-$function. All of these results will be deduced from an analysis of a problem in the variable satisfied by the probability P}}^y Also interest is total number $N$ of particles killed at $0$. For supercritical branching motion with particle drift and killing at $0$, Maillard [@M13] *et [@BBHM15] have, under hypotheses concerning the drift the mechanism, obtained sharp estimates tail the For branching motion with killing, T. Lee [@L90-1

$y>0$, $${{\mathbb P}}^{y} \{M\geq x \} \leq \frac{(6+\varePsilon)}{\sigmA^{2}x^{2}}$$ for All SufFiCienTly lArge $x$. It is by no mEAns eVident, however, that the taIl behAvIOr shOUlD be thE same as FOr BRAncHiNg broWnIAn MotioN wiTh no kilLing. In fact We wIlL prove that whEN tHe branchinG prOcess is initiAteD by a siNgLe pARticlE at A locaTion $y>0$ nEAr zero, The tail foLlOWs a powER law witH EXpOnenT $3$ rather than $2$. In partICuLAr, we will prove iN TheorEm \[THm.MAX.geNerAl\] that for eAcH fixeD $Y>0$, $${{\mathbb p}}^Y (M>X)\SIM \frAC{C_3y}{x^3} \quad \text{aS} \;\; x \rightarroW \InfTy.$$ wherE $C_{3}>0$ Is a COnstanT depeNdINg oN the offspriNg diStributioN but noT On $x$ or $y$. ON The otheR hand, wE wiLl sHow tHAt FoR inItIAl pARtIclE LocAtions $y=sX$ wHoSe disTancES FROm thE kiLlinG zone Are proportionAl tO the TArgEt $x$, thE expoNent Of The poWer law ReverTs To $2$; in particular, tHere Exists a coNtiNuOus FuNctioN $c_{4} (s)$ of $s\iN (0,1)$ suCh tHat $${{\mathBb P}}^{sx} (M>x)\SIm \fRaC{c_4 (S)}{X^2} \qUad \text{as} \;\; x \rightarrOw \INFtY.$$ FurtherMore, we WIlL sHOw that in ThE *MoRaniAN* Case, wHere THe OffsprinG law is DOuBlE-or-nothInG, the taIl ProBabIlity ${{\MAthbB P}}^y (M>x)$ cAn be explIcitlY Written as a WeieRStrass $\mathcal{p}-$FuNCTiON. All Of tHese results Will BE dedUced FRoM an ANalysIs of a BoUNdARy value problem in the VaRiable $Y$ satiSfied by the hitTing probabILITy ${{\mathbb p}}^y (M>x)$. aLsO Of interest is thE totaL number $N$ of PArticles KilleD at $0$. For suPercriticAL BranchinG BrOwnIan MotION wIth particle drIFT and KiLling at $0$, maiLlard [@M13] aNd BEreStyCki *Et Al.* [@BBHM15] havE, under vaRiOuS hYpOthEses cONcerning ThE drIfT anD the rEProducTion mEchaNiSm, OBtaIned shaRP eSTImatEs FoR the TaiL oF the dIstrIButIon of $N$. FOr criticaL brANchiNg brOwnian mOtion with killInG, T. Y. Lee [@L90-1

$y>0$, $${{\mathbb P}}^{y } \{M\geqx \}\le q \ fr ac{( 6+\v arepsilon)}{\s i gma^ {2}x^{2}}$$ for all su ffici en t ly l a rg e $x$ . It i s b y nome an s e vi d en t, ho wev er, tha t the tail be ha vior shouldb ethe same a s f or branching Br ownian m oti o n wit h n o kil ling.I n fact we willpr o ve tha t when t h e b ranc hing process is i n it i ated by a sing le par ti c le a t a lo cation $y> 0$ near zero, t h et a i l f o llows a power law with e x pon ent $3 $rat h er tha n $2$ .I n p articular,we w ill prove in Th e orem \[ t hm.max. genera l\] th at f o rea chfi x ed$ y> 0$, $${ {\mathbb P }} ^y (M >x)\ s i m \fra c{C _3y} {x^3} \quad \text{ as} \;\ ; x \ right arro w\inft y.$$ w here$C _{3}>0$ is a co nsta nt depend ing o n t he offs p ring d ist rib ution b ut noto n $ x$ o r $ y$. On the other h an d , w e will s how th a tfo r initial p art icle l ocati ons$ y= sx$ whos e dist a nc es from t he killi ng zo neare p r opor tional to thetarge t $x$, the expo n ent of the po w er l aw reve rts to $2$; in par t icul ar,t he ree xists a co nt i nu o us function $C_{4}(s )$ of$s\in (0,1)$ suchthat $${{\ m a t hbb P}}^ {sx} (M > x)\sim \frac{C _4 (s )}{x^2} \q u ad \text {as}\;\; x \right a r row \inf ty. $$Fur the r m or e, we will sh o w tha tin the*Mo ranian* ca se, wh ere t he offspr ing lawis d ou bl e-o r-not h ing, the t ail p rob abili t y ${{\ mathb b P} }^ y( M>x )$ canb ee x plic it ly wri tte nas aWeie r str ass $\m athcal{P} -$f u ncti on .All ofthese results w ill be ded uc edfrom a n analysis of a boundary value pr o blem in th e var iabl e $y$ sat isf ied by th e hitti ng pro babil it y $ { { \math b b P }}^ y(M>x)$. A l s o o f int er estis thetotal number $N$ o f pa rticles kille d a t $0 $ . F ors up e rcr it i cal b ranching Browni an motionwi t hparticle d r ift a nd kill ing at$0$,M aillard [@M13] a nd Berest yc ki * e t al .* [@BBHM1 5] have, under va r ioush yp othes esconcer ni ngthe d rift a n d t he re produc ti on mec hanis m, obtaine d sharp estimates for t he tai l ofthe distribu tio n of $N$. For cri tical bran chi ngBrown ian motio n wi t hkil l ing,T. Y .  Lee [@L9 0 -1

$y>0$,_$${{\mathbb P}}^{y}_\{M\geq x \} \leq_\frac{(6+\varepsilon)}{\sigma^{2}x^{2}}$$ for_all_sufficiently large_$x$. It_is by no_means evident, however,_that the tail behavior_should be the_same_as for branching Brownian motion with no killing. In fact we will prove that_when_the branching_process_is_initiated by a single particle_at a location $y>0$ near_zero, the_tail follows a power law with exponent $3$_rather_than $2$. In_particular, we will prove in Theorem \[thm.max.general\] that for_each fixed $y>0$, $${{\mathbb P}}^y (M>x)\sim_\frac{C_3y}{x^3} \quad \text{as}_\;\;_x _ \rightarrow_\infty.$$ where $C_{3}>0$ is a constant_depending on the offspring distribution but_not on $x$ or $y$. On the_other hand, we will show that_for initial particle locations $y=sx$_whose distances_from the killing zone are_proportional to the_target $x$,_the exponent of_the power law reverts to $2$;_in particular, there_exists a continuous function $C_{4} (s)$_of_$s\in (0,1)$ such_that_$${{\mathbb_P}}^{sx} (M>x)\sim_\frac{C_4 (s)}{x^2} \quad_\text{as}_\;\; x __ \rightarrow \infty.$$ Furthermore, we_will_show that in the *Moranian* case, where_the offspring law is_double-or-nothing,_the tail probability ${{\mathbb_P}}^y (M>x)$ can be explicitly written_as a Weierstrass $\mathcal{P}-$function. All of_these results_will be_deduced from an analysis of a boundary value problem in the_variable $y$ satisfied by the hitting_probability ${{\mathbb P}}^y (M>x)$. Also_of interest_is_the total number_$N$_of particles_killed at $0$. For supercritical branching Brownian_motion with_particle drift and killing at $0$,_Maillard [@M13] and Berestycki_*et_al.* [@BBHM15] have, under various hypotheses_concerning the drift and the reproduction_mechanism, obtained sharp estimates for_the_tail_of the distribution of $N$._For critical branching Brownian motion with_killing, T. Y. Lee [@L90-1

651\ \textrm{ with the 95\% confidence interval } \ [ 0.0035873 ; 0.0083429]$$ Although this value seems to be small, the significance of $ \hat\beta$ has been shown by a likelihood ratio test. The test of $\beta=0$ against $\beta=\hat\beta$ returns a p-value of $7.02 \times 10^{-7}$. Thus, the solar activity index $X$ affects the number of storms occurring during a cycle. Graphically, the influence of the solar activity index on the number of storms per cycle is observable on Figure \[NbO\_AS\_111\].\ ![Total number of storms per cycle for a low level of 111 according to the solar activity (centered)[]{data-label="NbO_AS_111"}](Fig5.jpg){width="8cm" height="6.5cm"} Instantaneous intensity: extrapolation to level 400 and relative risk {#section : extrapol} --------------------------------------------------------------------- The extrapolation to the storms of extreme level 400 is made by multiplying by $\widehat{P_{400}}$ (with confidence interval). We obtain the final intensity shown in Figure \[fig: intensite\_111\]. This curve corresponds to intensity of apparition of extreme storms for a solar cycle with a mean solar activity of 146.7. Recall that in the equation: $$\lambda_j(t) = \lambda_0(t) D_j \exp ( \beta X_j)$$ the risk factor is $\exp ( \beta X_j)$. Then, using $\hat\beta$, we can evaluate the relative risk for a cycle with a given solar activity index. For example compared to the average level of solar activity (146.7), a cycle with a high solar activity of 180 has a relative risk of $\exp(33.3 \times 0.0059651) = 1.22$. ![Instantaneous intensity (years$^{-1}$), with confidence interval, of the storms of level 400 obtained by extrapolation from the low level 111, for a mean solar activity of 146.7. In dash-dotted line the empirical frequency of storms of level 400[]{data-label="fig: intensite_111"}](Fig6.jpg){width="8cm" height="6.5cm"} Method stability ---------------- The results presented in the previous sections are given for a fixed low level (of 111). This asks the question

651\ \textrm { with the 95\% confidence interval } \ [ 0.0035873 ; 0.0083429]$$ Although this value seems to be small, the meaning of $ \hat\beta$ has been usher by a likelihood ratio test. The trial of $ \beta=0 $ against $ \beta=\hat\beta$ return a p - value of $ 7.02 \times 10^{-7}$. Thus, the solar activeness exponent $ X$ affects the number of storms occurring during a cycle. diagrammatically, the influence of the solar activity index on the number of storms per hertz is observable on Figure \[NbO\_AS\_111\].\ ! [ Total number of storms per cycle for a abject level of 111 according to the solar activity (centered)[]{data - label="NbO_AS_111"}](Fig5.jpg){width="8 curium " height="6.5 cm " } Instantaneous intensity: extrapolation to level 400 and relative risk { # part: extrapol } --------------------------------------------------------------------- The extrapolation to the storms of extreme level 400 is made by multiplying by $ \widehat{P_{400}}$ (with confidence time interval). We obtain the final intensity shown in Figure \[fig: intensite\_111\ ]. This curve corresponds to volume of apparition of extreme storms for a solar cycle with a mean solar activity of 146.7. Recall that in the equation: $ $ \lambda_j(t) = \lambda_0(t) D_j \exp (\beta X_j)$$ the risk factor is $ \exp (\beta X_j)$. Then, using $ \hat\beta$, we can evaluate the proportional hazard for a cycle with a given solar activity exponent. For exercise compared to the average grade of solar natural process (146.7), a cycle with a high solar natural process of 180 has a relative risk of $ \exp(33.3 \times 0.0059651) = 1.22$. ! [ Instantaneous intensity (years$^{-1}$), with confidence interval, of the storm of level 400 obtained by extrapolation from the low level 111, for a hateful solar action of 146.7. In dash - dotted line the empirical frequency of storms of floor 400[]{data - label="fig: intensite_111"}](Fig6.jpg){width="8 cm " height="6.5 cm " } Method stability ---------------- The results presented in the previous sections are given for a fixed low horizontal surface (of 111). This asks the motion

651\ \tedtrm{ with the 95\% confidenct interval } \ [ 0.0035873 ; 0.0083429]$$ Althoujh this value sdems to be small, the signifirancw of $ \hat\beta$ has been showv by a linelihood ratmo test. The test of $\beta=0$ againab $\betc=\het\beta$ returns s p-value ox $7.02 \times 10^{-7}$. Thus, tfe solar activity index $X$ affects the number ov storms occurting ctrinf a cycle. Graphically, the influencs of tht solar activity imdex on the number of storls pfr cycle is observwble on Figode \[GvO\_AS\_111\].\ ![Total nuober of storms per cycme for a low level of 111 accordine to che solar axtuvihi (centered)[]{dava-labej="NbO_AS_111"}](Fig5.jpg){width="8cm" veight="6.5ck"} Instantaneous invensuty: extrapolation to nevel 400 and relativg risk {#secdikn : extrapol} --------------------------------------------------------------------- The ezteapoldtiot to rhe stkrks of exhreje level 400 js made by nultiplying by $\widenae{I_{400}}$ (with confidsnce igtqrval). We obtain the final intensity shofn jn Figure \[fig: intensite\_111\]. This curve corresponfs to intqnsity of apparition of extreme storms for a solas cycmd wnbm a owaj solar activity of 146.7. Recall that in the equatykn: $$\lsmbda_j(t) = \lambdc_0(t) D_j \exp ( \beta C_j)$$ tng risk factor ks $\exp ( \bsta X_j)$. Then, using $\jat\beta$, we cqn evaluaue thr relative risk for a cycle with a giveu silar activity inder. For examplz compsred yo the average level of solzr activity (146.7), a cycle dith a high solaf abtivhty of 180 has a relative rish of $\exp(33.3 \vimes 0.0059651) = 1.22$. ![Instxntameous yntensity (jears$^{-1}$), with confidence inhervap, mf the stogms of level 400 obtained by extra'plation from tve now levej 111, fov a mean solar wctivity of 146.7. Iu dash-docted lkne the emiirical fcequency of ftorms of levap 400[]{data-label="fmg: intensyte_111"}](Fug6.jpt){width="8co" height="6.5cm"} Methoc stabiliny ---------------- Uhe resultw presented in the prgvjous sections axt guven for a fixec luw jenel (os 111). This asks dhe dueryion

651\ \textrm{ with the 95\% confidence interval [ ; 0.0083429]$$ this value seems of \hat\beta$ has been by a likelihood test. The test of $\beta=0$ against returns a p-value of $7.02 \times 10^{-7}$. Thus, the solar activity index $X$ the number of storms occurring during a cycle. Graphically, the influence of the activity on number storms per cycle is observable on Figure \[NbO\_AS\_111\].\ ![Total number of storms per cycle for a level of 111 according to the solar activity height="6.5cm"} Instantaneous intensity: extrapolation level 400 and relative risk : --------------------------------------------------------------------- The to storms extreme level 400 made by multiplying by $\widehat{P_{400}}$ (with confidence interval). We obtain the final intensity shown in Figure \[fig: This curve intensity of of storms a solar cycle mean solar activity of 146.7. Recall equation: $$\lambda_j(t) = \lambda_0(t) D_j \exp ( \beta the risk is $\exp ( \beta X_j)$. Then, $\hat\beta$, we can evaluate the relative risk for cycle with a given solar activity index. For example compared to the average level of (146.7), a cycle with high solar activity 180 a risk $\exp(33.3 \times = 1.22$. ![Instantaneous intensity (years$^{-1}$), with confidence interval, of the storms level 400 obtained by extrapolation from the low level 111, mean activity of 146.7. dash-dotted line the empirical of of level 400[]{data-label="fig: intensite_111"}](Fig6.jpg){width="8cm" stability The the sections given for a fixed level (of 111). This asks question

651\ \textrm{ with the 95\% confidence inTerval } \ [ 0.0035873 ; 0.0083429]$$ AlthOugh tHis ValUe SeemS to bE small, the signiFIcanCe of $ \hat\beta$ has been showN by a lIkELihoOD rAtio tEst. The tESt OF $\BetA=0$ aGaInsT $\bETa=\Hat\beTa$ rEturns a P-value of $7.02 \tiMes 10^{-7}$. thUs, the solar acTIvIty index $X$ aFfeCts the number Of sTorms oCcUrrINg durIng A cyclE. GraphICally, tHe influenCe OF the soLAr activITY iNdex On the number of storMS pER cycle is observAble on fiGUrE \[nBO\_As\_111\].\ ![ToTal number oF sTorms PEr cycle FOr A LOW leVEl of 111 according To the solar aCTivIty (cenTeRed)[]{DAta-labEl="NbO_aS_111"}](fIg5.jPg){width="8cm" heIght="6.5Cm"} InstantAneous INtensitY: ExtrapoLation To lEveL 400 and RElAtIve RiSK {#seCTiOn : eXTraPol} --------------------------------------------------------------------- The exTrApOlatiOn to THE STormS of ExtrEme leVel 400 is made by muLtiPlyiNG by $\WidehAt{P_{400}}$ (wiTh coNfIdencE interVal). We ObTain the final intEnsiTy shown in figUrE \[fiG: iNtensITe\_111\]. This CurVe cOrrespoNds to inTEnsItY OF ApParition of extreme sToRMS fOr a solar Cycle wITh A mEAn solar aCtIviTy of 146.7. rECall tHat iN ThE equatioN: $$\lambdA_J(t) = \LaMbda_0(t) D_j \ExP ( \beta X_J)$$ tHe rIsk FactoR Is $\exP ( \beta X_J)$. Then, usiNg $\hat\BEta$, we can evaluaTE the relative rISk FOR a CYcle WitH a given solaR actIVity IndeX. foR exAMple cOmparEd TO tHE average level of solaR aCtivitY (146.7), a cycLe with a high soLar activitY OF 180 Has a relaTive RIsK Of $\exp(33.3 \times 0.0059651) = 1.22$. ![InstAntanEous intensITy (years$^{-1}$), wIth coNfidence Interval, oF THe storms Of lEveL 400 obTaiNED bY extrapolatioN FRom tHe Low leveL 111, foR a mean sOlaR acTivIty Of 146.7. in dash-dotTed line tHe EmPiRiCal FrequENcy of stoRmS of LeVel 400[]{Data-lABel="fig: IntenSite_111"}](fiG6.jPG){wiDth="8cm" heIGhT="6.5CM"} MetHoD sTabiLitY ---------------- THe resUlts PResEnted in The previoUs sECtioNs ArE given fOr a fixed low leVeL (of 111). This askS tHe qUestioN

651\ \textrm{ with the 95\ % confiden ce in ter val } \ [ 0.0 035873 ; 0.00 8 3429 ]$$ Although this valu e see ms to b e s mall, the si g ni f i can ce o f $ \ h at \beta $ h as been shown bya l ik elihood rati o t est. The t est of $\beta=0 $ a gainst $ \be t a=\ha t\b eta$return s a p-v alue of $ 7. 0 2 \tim e s 10^{- 7 } $. Thu s, the solar acti v it y index $X$ aff ects t he nu m b erofstorms occ ur ringd uring a cy c l e . G r aphically, th e influence ofthe so la r a c tivity inde xo n t he number o f st orms percyclei s obser v able on Figur e \ [Nb O\_A S \_ 11 1\] .\ ![ T ot aln umb er of st or ms percycl e f o r alow lev el of 111 accordin g t o th e so lar a ctivi ty ( ce ntere d)[]{d ata-l ab el="NbO_AS_111" }](F ig5.jpg){ wid th ="8 cm " hei g ht="6. 5cm "} Instan taneous int en s i t y: extrapolation tole v e l400 andrelati v eri s k {#sect io n : ext r a pol}---- - -- -------- ------ - -- -- ------- -- ------ -- --- --- ----- - ---- ------ ---- Th e ext r apolation to t h e storms of e x tr e m el evel 40 0 is made b y mu l tipl ying by $\ w ideha t{P_{ 40 0 }} $ (with confidence i nt erval) . Weobtain the fi nal intens i t y shown i n Fi g ur e \[fig: intens ite\_ 111\]. Thi s curve c orres ponds to intensit y of appar iti onofext r e me storms for a s olar c ycle wi tha meansol aract ivi ty of 146.7 . Recall t ha tin th e equ a tion: $$ \l amb da _j( t) =\ lambda _0(t) D_j \ ex p (\beta X _ j) $ $ the r is k fa cto ris $\ exp( \b eta X_j )$. Then, us i ng $ \h at \beta$, we can evalu at e the rela ti verisk f o r a cycle with a given solar act i vity in dex . For exa mple comp are d to t hea verage level of s ol ara c tivit y (1 46. 7) , a cyclew i tha hig hsola r activ ity of 180 has a r e lat ive risk of $ \ex p(33 . 3 \ tim e s0 .00 59 6 51) = 1.22$. ![Inst antaneousin t en sity (year s $^{ -1 }$), wi th conf idenc e interv al, of th e stormsof lev e l 40 0 obtained by extr apolation fromt he lowlev el 111 ,for a me an sol a r a ctivi ty of14 6.7. I n das h- dotted l ine the empirical frequ ency o f sto rms of level 40 0 []{ data-labe l="f ig: intens ite _11 1"}]( Fig 6 .jpg) {wid t h= "8c m " hei ght= " 6.5cm"} M et hod s ta bility ---- - - - --- ----- - T he res ults presented in the previous secti onsa r e g ive n for a fixed low lev el(o f 111). Th is asks the q uestion

651\ \textrm{_with the_95\% confidence interval }_\ [_0.0035873_ ;_0.0083429]$$_Although this value_seems to be_small, the significance of_$ \hat\beta$ has_been_shown by a likelihood ratio test. The test of $\beta=0$ against $\beta=\hat\beta$ returns a_p-value_of $7.02_\times_10^{-7}$._Thus, the solar activity index_$X$ affects the number of_storms occurring_during a cycle. Graphically, the influence of the_solar_activity index on_the number of storms per cycle is observable on_Figure \[NbO\_AS\_111\].\ ![Total number of storms per_cycle for a_low_level_of 111 according to_the solar activity (centered)[]{data-label="NbO_AS_111"}](Fig5.jpg){width="8cm" height="6.5cm"} Instantaneous intensity:_extrapolation to level 400 and relative_risk {#section : extrapol} --------------------------------------------------------------------- The extrapolation to the_storms of extreme level 400 is_made by multiplying by $\widehat{P_{400}}$_(with confidence_interval). We obtain the final_intensity shown in_Figure \[fig:_intensite\_111\]. This curve_corresponds to intensity of apparition of_extreme storms for_a solar cycle with a mean_solar_activity of 146.7._Recall_that_in the_equation: $$\lambda_j(t) =_\lambda_0(t)_D_j \exp_(_\beta X_j)$$ the risk factor is_$\exp_( \beta X_j)$. Then, using $\hat\beta$, we_can evaluate the relative_risk_for a cycle with_a given solar activity index._For example compared to the average_level of_solar activity_(146.7), a cycle with a high solar activity of 180 has_a relative risk of $\exp(33.3 \times_0.0059651) = 1.22$. ![Instantaneous intensity_(years$^{-1}$), with_confidence_interval, of the_storms_of level_400 obtained by extrapolation from the low_level 111,_for a mean solar activity of_146.7. In dash-dotted line_the_empirical frequency of storms of level_400[]{data-label="fig: intensite_111"}](Fig6.jpg){width="8cm" height="6.5cm"} Method stability ---------------- The results presented_in the previous sections are_given_for_a fixed low level (of_111). This asks the question

and S. Rakshit, hep-ph/0102199, S. Komine, T. Moroi and M. Yamaguchi, hep-ph/0102204, S.N. Gninenko and N.V. Krasnikov, hep-ph/0102222, K. Cheung, hep-ph/0102238, P. Das, S. K. Rai and S. Raychaudhuri, hep-ph/0102242, T.W. Kephart, H. Päs, hep-ph/0102243, E. Ma and M. Raidal, hep-ph/0102255, Z. Xiong and J. M. Yang, hep-ph/0102259, A. Dedes and H .E. Haber, hep-ph/0102297, Z. Xing, hep-ph/0102304, J. Hisano and K. Tobe, hep-ph/0102315, T. Ibrahim, U. Chattopadhyay and P. Nath, hep-ph/0102324, J. Ellis, D. V. Nanopoulos and K. A. Olive, hep-ph/0102331, R. Arnowitt, B. Dutta, B. Hu and Y. Santoso, hep-ph/0102344. S. J. Brodsky and S. D. Drell, Phys. Rev. D [**22**]{}, 2236 (1980). H. Fritzsch and D. Holtmannspotter, Phys. Lett. B [**457**]{}, 186 (1999) \[hep-ph/9901411\]. H. Fritzsch and D. Holtmannspotter, Phys. Lett. B [**338**]{}, 290 (1994) \[hep-ph/9406241\]. D. E. Groom [*et al.*]{} \[Particle Data Group Collaboration\], Eur. Phys. J. C [**15**]{}, 1 (2000). --- abstract: 'The properties of a Bose gas of translation-invariant (TI) bipolarons analogous to Cooper pairs are considered. As in the BCS theory, the description of a

and S.   Rakshit, hep - ph/0102199, S.   Komine, T.   Moroi and M.   Yamaguchi, hep - ph/0102204, S.N. Gninenko and N.V. Krasnikov, hep - ph/0102222, K.   Cheung, hep - ph/0102238, P.   Das, S.   K.   Rai and S.   Raychaudhuri, hep - ph/0102242, T.W. Kephart, H. Päs, hep - ph/0102243, E.   Ma and M.   Raidal, hep - ph/0102255, Z.   Xiong and J. M. Yang, hep - ph/0102259, A.   Dedes and H   .E.   Haber, hep - ph/0102297, Z.   Xing, hep - ph/0102304, J.   Hisano and K.   Tobe, hep - ph/0102315, T.   Ibrahim, U.   Chattopadhyay and P.   Nath, hep - ph/0102324, J.   Ellis, D.   V.   Nanopoulos and K.   A.   Olive, hep - ph/0102331, R.   Arnowitt, B.   Dutta, B.   Hu and Y.   Santoso, hep - ph/0102344. S.   J.   Brodsky and S.   D.   Drell, Phys.   Rev.   D [ * * 22 * * ] { }, 2236 (1980). H.   Fritzsch and D.   Holtmannspotter, Phys.   Lett.   B [ * * 457 * * ] { }, 186 (1999) \[hep - ph/9901411\ ]. H.   Fritzsch and D.   Holtmannspotter, Phys.   Lett.   B [ * * 338 * * ] { }, 290 (1994) \[hep - ph/9406241\ ]. D.   E.   Groom [ * et al. * ] { } \[Particle Data Group Collaboration\ ], Eur.   Phys.   J.   C [ * * 15 * * ] { }, 1 (2000). --- abstract:' The properties of a Bose gas of translation - invariant (TI) bipolarons analogous to Cooper pair are view. As in the BCS theory, the description of a

anf S. Rakshit, hep-ph/0102199, S. Komine, T. Moroi and M. Yamaguchi, hep-ph/0102204, S.N. Gnindnko and N.V. Krasnikov, hep-ph/0102222, N. Cyeung, hep-ph/0102238, P. Das, S. K. Rai and R. Raychaudjuri, hep-ph/0102242, T.X. Kephart, H. Päs, hxl-ph/0102243, E. Ma and M. Dwidan, hep-ph/0102255, Z. Xiong snd J. M. Yatg, hep-ph/0102259, A. Dedev xnb H .E. Haber, hep-ph/0102297, Z. Xing, hep-ph/0102304, J. Hisano wnd K. Tone, hep-ph/0102315, T. Ibrahii, U. Cnwttoladhyay and P. Nath, hep-ph/0102324, J. Ellis, D. V. Nznopoulms and K. A. Olivr, hep-ph/0102331, R. Arnowitt, B. Dutta, B. Ju ajd Y. Santoso, hep-ph/0102344. D. J. Brodsky abd S. Q. Erell, Phys. Rex. D [**22**]{}, 2236 (1980). H. Friuzvch and D. Hkltmannspotter, Phys. Lett. B [**457**]{}, 186 (1999) \[hep-oh/9901411\]. H. Fxitzsch and D. Yolhkannspotter, Phys. Jett. B [**338**]{}, 290 (1994) \[hep-in/9406241\]. D. E. Grmom [*et sl.*]{} \[Particle Daba Grmup Collaboration\], Eur. Phyv. J. C [**15**]{}, 1 (2000). --- abstract: 'Thg propertias of a Bose gas of teansldtiot-invxeiavt (UI) uipklarond aialogous to Cooper paies are considered. Ax yb the BCS thekry, thq qescription of a

and S. Rakshit, hep-ph/0102199, S. Komine, T. M. hep-ph/0102204, S.N. and N.V. Krasnikov, Das, K. Rai and Raychaudhuri, hep-ph/0102242, T.W. H. Päs, hep-ph/0102243, E. Ma and Raidal, hep-ph/0102255, Z. Xiong and J. M. Yang, hep-ph/0102259, A. Dedes and H Haber, hep-ph/0102297, Z. Xing, hep-ph/0102304, J. Hisano and K. Tobe, hep-ph/0102315, T. Ibrahim, Chattopadhyay P. hep-ph/0102324, Ellis, D. V. Nanopoulos and K. A. Olive, hep-ph/0102331, R. Arnowitt, B. Dutta, B. Hu and Santoso, hep-ph/0102344. S. J. Brodsky and S. D. Phys. Rev. D [**22**]{}, (1980). H. Fritzsch and D. Phys. B [**457**]{}, (1999) H. and D. Holtmannspotter, Lett. B [**338**]{}, 290 (1994) \[hep-ph/9406241\]. D. E. Groom [*et al.*]{} \[Particle Data Group Collaboration\], Eur. Phys. C [**15**]{}, --- abstract: properties a gas of translation-invariant analogous to Cooper pairs are considered. BCS theory, the description of a

and S. Rakshit, hep-ph/0102199, S. Komine, T. MOroi and M. YaMagucHi, hEp-pH/0102204, S.n. GniNenkO and N.V. KrasnikoV, Hep-pH/0102222, K. Cheung, hep-ph/0102238, P. Das, S. K. Rai And S. RAyCHaudHUrI, hep-pH/0102242, T.W. KephARt, h. pÄs, hEp-Ph/0102243, e. Ma AnD m. RAidal, Hep-Ph/0102255, Z. XionG and J. M. Yang, Hep-Ph/0102259, a. Dedes and H .E. HABeR, hep-ph/0102297, Z. XinG, heP-ph/0102304, J. Hisano anD K. TObe, hep-Ph/0102315, t. IbRAhim, U. chaTtopaDhyay aND P. Nath, Hep-ph/0102324, J. EllIs, d. v. NanopOUlos and k. a. olIve, hEp-ph/0102331, R. Arnowitt, B. DutTA, B. hU and Y. Santoso, heP-ph/0102344. S. J. BRoDSkY ANd S. d. DrEll, Phys. Rev. d [**22**]{}, 2236 (1980). H. fritzSCh and D. HOLtMANNspOTter, Phys. Lett. B [**457**]{}, 186 (1999) \[Hep-ph/9901411\]. H. FritzSCh aNd D. HolTmAnnSPotter, phys. LEtT. b [**338**]{}, 290 (1994) \[heP-ph/9406241\]. D. E. Groom [*eT al.*]{} \[PArticle DaTa GrouP collaboRAtion\], EuR. Phys. J. c [**15**]{}, 1 (2000). --- abStrAct: 'THE pRoPerTiES of A boSe gAS of TranslatIoN-iNvariAnt (Ti) BIPOlarOns AnalOgous To Cooper pairs Are ConsIDerEd. As iN the BcS thEoRy, the DescriPtion Of A

and S. Rakshit, hep-ph/01 02199, S.Komin e,T.Mo roiandM. Yamaguchi,h ep-p h/0102204, S.N. Gninen ko an dN .V.K ra sniko v, hep- p h/ 0 1 022 22 ,K.Ch e un g, he p-p h/01022 38, P. Das , S .K. Rai and S .  R aychaudhur i,hep-ph/01022 42, T.W.Ke pha r t, H. Pä s, he p-ph/0 1 02243, E. Ma an dM . Raid a l, hep- p h /0 1022 55, Z. Xiong andJ .M . Yang, hep-ph /01022 59 , A . Ded esand H .E.Ha ber,h ep-ph/0 1 02 2 9 7 , Z .  Xing, hep-ph /0102304, J .  Hi sano a nd K. Tobe,hep-p h/ 0 102 315, T. Ibr ahim , U. Chat topadh y ay andP . Nath, hep-p h/0 102 324, J.  E lli s, D.V .Nan o pou los andK.  A . Oli ve,h e p - ph/0 102 331, R. A rnowitt, B. D utt a, B .  Hu andY. Sa ntos o, hep- ph/010 2344. S . J. Brodsky an d S.  D. Drell , P hy s.Re v. D[ **22** ]{} , 2 236 (19 80). H. Fri tz s c h a nd D. Holtmannspot te r , P hys. Let t. B [ * *4 57 * *]{}, 18 6(19 99)\ [ hep-p h/99 0 14 11\]. H.  Fritz s ch a nd D. H ol tmanns po tte r,Phys. Lett . B [* *338**]{ }, 29 0 (1994) \[hep- p h/9406241\].D .E .  G r oom[*e t al.*]{} \ [Par t icle Dat a G rou p Coll abora ti o n\ ] , Eur. Phys. J. C [ ** 15**]{ }, 1(2000). ---abstract:' T h e proper ties of a Bose gas oftrans lation-inv a riant (T I) bi polarons analogou s to Coope r p air s a rec o ns idered. As in t he B CS theory , t he desc rip tio n o f a

and_S. Rakshit, hep-ph/0102199,_S. Komine, T. Moroi and M. Yamaguchi,_hep-ph/0102204, S.N._Gninenko_and N.V._Krasnikov,_hep-ph/0102222, K. Cheung, hep-ph/0102238,_P. Das, S. K. Rai and_S. Raychaudhuri, hep-ph/0102242, T.W. Kephart,_H. Päs, hep-ph/0102243,_E. Ma_and M. Raidal, hep-ph/0102255, Z. Xiong and J. M. Yang, hep-ph/0102259, A. Dedes and H .E. Haber, hep-ph/0102297, Z. Xing,_hep-ph/0102304,_J. Hisano and_K. Tobe,_hep-ph/0102315,_T. Ibrahim, U. Chattopadhyay and P. Nath, hep-ph/0102324,_J. Ellis, D. V. Nanopoulos and K. A. Olive, hep-ph/0102331,_R. Arnowitt, B. Dutta,_B. Hu and Y. Santoso, hep-ph/0102344. S. J. Brodsky and S. D. Drell, Phys. Rev. D_[**22**]{},_2236 (1980). H. Fritzsch_and D. Holtmannspotter, Phys. Lett. B [**457**]{}, 186 (1999) \[hep-ph/9901411\]. H. Fritzsch and_D. Holtmannspotter, Phys. Lett. B [**338**]{}, 290 (1994) \[hep-ph/9406241\]._D. E. Groom [*et al.*]{}_\[Particle_Data_Group Collaboration\], Eur. Phys. J. C [**15**]{},_1 (2000). --- abstract: 'The properties of_a Bose gas of translation-invariant (TI)_bipolarons analogous to Cooper pairs are considered._As in the BCS theory, the_description of a

8z) \in (0, 1]$ in the preceding inequality, rearrangement yields the desired estimate. Our proof of also verifies that the operator $\Delta_A$ in is bounded since $\Delta_{A_s}$ is bounded with the same domain and range spaces. Next, we have the The finite-dimensionality of $\Ker\Delta_A\cap W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P)$ follows from Corollary \[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\]. We observe that, by increasing the constant $C$ as needed, the term $\|\xi\|_{L^p(X)}$ appearing on the right-hand side of the inequality can be replaced by $\|\xi\|_{L^2(X)}$. This is clear when $p \leq 2$, while if $p > 2$, we can choose $s \in (p,\infty)$ and apply the interpolation inequality [@GilbargTrudinger Equation (7.10)], $$\|\xi\|_{L^p(X)} \leq \delta \|\xi\|_{L^s(X)} + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ for $\nu := (1/2-1/p)/(1/p-1/s) > 0$ and arbitrary positive $\delta$. Because $p \geq d/2$, we have a continuous Sobolev embedding $W^{2,p}(X) \subset L^s(X)$ as already observed, so $$\|\xi\|_{L^p(X)} \leq C_1\delta \|\xi\|_{W_{A_1}^{2,p}(X)} + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ where $C_1 = C_1(A_1,g,l,p) \in [1,\infty)$. Hence, for $\delta(A_1,g,l,p) \in (0,1]$ given by $\delta = 1/(2CC_1)$, we can use rearrangement in to replace $\|\xi\|_{L^p(X)}$ by $\|\xi\|_{L^2(X)}$. Therefore, the estimate implies $$\label{eq:W2p_apriori_estimate_Delta_A_Sobolev_L2_xi_rhs} \|\xi\|_{

8z) \in (0, 1]$ in the preceding inequality, rearrangement yields the desired estimate. Our validation of besides verifies that the hustler $ \Delta_A$ in is bounded since $ \Delta_{A_s}$ is bounded with the same world and range spaces. Next, we hold the The finite - dimensionality of $ \Ker\Delta_A\cap W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P)$ follows from Corollary \[cor: Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\ ]. We observe that, by increase the constant $ C$ as needed, the terminus $ \|\xi\|_{L^p(X)}$ appearing on the right - handwriting slope of the inequality can be replaced by $ \|\xi\|_{L^2(X)}$. This is clear when $ p \leq 2 $, while if $ p > 2 $, we can choose $ s \in (p,\infty)$ and apply the interpolation inequality [ @GilbargTrudinger Equation (7.10) ], $ $ \|\xi\|_{L^p(X) } \leq \delta \|\xi\|_{L^s(X) } + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ for $ \nu: = (1/2 - 1 / p)/(1 / p-1 / s) > 0 $ and arbitrary convinced $ \delta$. Because $ p \geq d/2 $, we have a continuous Sobolev embedding $ W^{2,p}(X) \subset L^s(X)$ as already observe, so $ $ \|\xi\|_{L^p(X) } \leq C_1\delta \|\xi\|_{W_{A_1}^{2,p}(X) } + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ where $ C_1 = C_1(A_1,g, l, p) \in [ 1,\infty)$. Hence, for $ \delta(A_1,g, l, p) \in (0,1]$ given by $ \delta = 1/(2CC_1)$, we can use rearrangement in to supplant $ \|\xi\|_{L^p(X)}$ by $ \|\xi\|_{L^2(X)}$. Therefore, the estimate implies $ $ \label{eq: W2p_apriori_estimate_Delta_A_Sobolev_L2_xi_rhs } \|\xi\| _ {

8z) \ij (0, 1]$ in the preceding ineduality, rearranywment bields fhe desifed estimate. Our proof of alsl cerifues that the operator $\Aelta_A$ in is bouneed wunce $\Delta_{E_a}$ is bomuded slth tke same domain akd range spdces. Next, we haee tke The finite-dimensionality of $\Ker\Delea_A\cap E_{A_1}^{2,o}(X;\Lambda^l\otimef\ad K)$ fjlloss from Corollary \[cor:Fredholmness\_ahd\_index\_Naplace\_operatpr\_on\_W2p\_Sobolev\_connection\]. We obsfrve that, by increwsing the cinstwbt $C$ as needdd, the term $\|\xi\|_{L^p(X)}$ appezring on the right-hand side of ghe iuequality cqn be teplaced by $\|\ei\|_{L^2(X)}$. Tris is clear when $p \neq 2$, whole if $p > 2$, we gan cioosw $s \in (p,\infty)$ and appny the interpolatijn inequanicy [@GilbargTrudinger Ewuqtion (7.10)], $$\|\xi\|_{N^p(X)} \uwq \aelua \|\ei\|_{L^a(X)} + \depta^{-\iu}\|\xi\|_{L^2(X)},$$ for $\hu := (1/2-1/p)/(1/p-1/s) > 0$ abd arbitrary positife $\eelta$. Because $p \geq d/2$, we have a continuous Sobolev embedding $W^{2,p}(S) \subset L^s(X)$ as already observed, so $$\|\xi\|_{L^p(X)} \lee C_1\delta \|\vi\|_{W_{A_1}^{2,p}(X)} + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ where $C_1 = C_1(A_1,g,l,p) \in [1,\infty)$. Hence, for $\veuta(C_1,n,j,o) \ij (0,1]$ given by $\delta = 1/(2CC_1)$, we can use rearrangemene im no replace $\|\xi\|_{L^p(X)}$ ny $\|\xi\|_{L^2(X)}$. Therefore, yhf rftimate impligs $$\label{eq:S2p_apriori_estimate_Dflta_A_Sofolev_O2_xi_rhs} \|\xi\|_{

8z) \in (0, 1]$ in the preceding yields desired estimate. proof of also in bounded since $\Delta_{A_s}$ bounded with the domain and range spaces. Next, we the The finite-dimensionality of $\Ker\Delta_A\cap W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad P)$ follows from Corollary \[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\]. We observe by increasing the constant $C$ as needed, the term $\|\xi\|_{L^p(X)}$ appearing on the side the can replaced by $\|\xi\|_{L^2(X)}$. This is clear when $p \leq 2$, while if $p > 2$, we choose $s \in (p,\infty)$ and apply the interpolation [@GilbargTrudinger Equation (7.10)], $$\|\xi\|_{L^p(X)} \delta \|\xi\|_{L^s(X)} + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ for := > 0$ arbitrary $\delta$. $p \geq d/2$, have a continuous Sobolev embedding $W^{2,p}(X) \subset L^s(X)$ as already observed, so $$\|\xi\|_{L^p(X)} \leq C_1\delta \|\xi\|_{W_{A_1}^{2,p}(X)} + where $C_1 \in [1,\infty)$. for \in given by $\delta we can use rearrangement in to $\|\xi\|_{L^2(X)}$. Therefore, the estimate implies $$\label{eq:W2p_apriori_estimate_Delta_A_Sobolev_L2_xi_rhs} \|\xi\|_{

8z) \in (0, 1]$ in the preceding inequaliTy, rearrangEment YieLds ThE desIred Estimate. Our proOF of aLso verifies that the operAtor $\DElTA_A$ in IS bOundeD since $\DELtA_{a_S}$ is BoUnDed WiTH tHe samE doMain and Range spaceS. NeXt, We have the The FInIte-dimensiOnaLity of $\Ker\DelTa_A\Cap W_{A_1}^{2,p}(x;\LAmbDA^l\otiMes\Ad P)$ foLlows fROm CoroLlary \[cor:FReDHolmneSS\_and\_indEX\_laPlacE\_operator\_on\_W2p\_SoboLEv\_COnnection\]. We obsErve thAt, BY iNCReaSinG the constaNt $c$ as neEDed, the tERm $\|\XI\|_{l^P(X)}$ aPPearing on the rIght-hand sidE Of tHe ineqUaLitY Can be rEplacEd BY $\|\xi\|_{l^2(X)}$. This is cleAr whEn $p \leq 2$, whiLe if $p > 2$, wE Can chooSE $s \in (p,\inFty)$ and AppLy tHe inTErPoLatIoN IneQUaLitY [@gilBargTrudInGeR EquaTion (7.10)], $$\|\XI\|_{l^P(x)} \leq \DelTa \|\xi\|_{l^s(X)} + \deLta^{-\nu}\|\xi\|_{L^2(X)},$$ for $\nU := (1/2-1/p)/(1/p-1/S) > 0$ and ARbiTrary PositIve $\dElTa$. BecAuse $p \gEq d/2$, we HaVe a continuous SoBoleV embeddinG $W^{2,p}(x) \sUbsEt l^s(X)$ as ALready ObsErvEd, so $$\|\xi\|_{L^P(X)} \leq C_1\dELta \|\Xi\|_{w_{a_1}^{2,P}(x)} + \dElta^{-\nu}\|\xi\|_{L^2(X)},$$ where $C_1 = C_1(A_1,G,l,P) \IN [1,\iNfty)$. HencE, for $\deLTa(a_1,g,L,P) \in (0,1]$ given By $\DelTa = 1/(2CC_1)$, WE Can usE reaRRaNgement iN to repLAcE $\|\xI\|_{L^p(X)}$ by $\|\xI\|_{L^2(x)}$. ThereFoRe, tHe eStimaTE impLies $$\laBel{eq:W2p_aPriorI_Estimate_Delta_A_sObolev_L2_xi_rhs} \|\xI\|_{

8z) \in (0, 1]$ in the pre ceding ine quali ty, re ar rang emen t yields the d e sire d estimate. Our proof of a ls o ver i fi es th at theo pe r a tor $ \D elt a_ A $in is bo unded s ince $\Del ta_ {A _s}$ is boun d ed with thesam e domain and ra nge sp ac es. Next , w e hav e theThe fi nite-dime ns i onalit y of $\K e r \D elta _A\cap W_{A_1}^{2 , p} ( X;\Lambda^l\ot imes\a dP )$ f oll ows from Coro ll ary \ [ cor:Fre d ho l m n ess \ _and\_index\_ Laplace\_op e rat or\_on \_ W2p \ _Sobol ev\_c on n ect ion\]. We o bser ve that,by inc r easingt he cons tant $ C$asneed e d, t hete r m $ \ |\ xi\ | _{L ^p(X)}$ap pe aring ont h e righ t-h andsideof the inequa lit y ca n be repl acedby $ \| \xi\| _{L^2( X)}$. T his is clear wh en $ p \leq 2$ , w hi leif $p > 2$, we ca n c hoose $ s \in ( p ,\i nf t y ) $and apply the inte rp o l at ion ineq uality [@ Gi l bargTrud in ger Equ a t ion ( 7.10 ) ], $$\|\xi \|_{L^ p (X )} \leq \ de lta \| \x i\| _{L ^s(X) } + \ delta^ {-\nu}\| \xi\| _ {L^2(X)},$$ fo r $\nu := (1/2 - 1/ p ) /( 1 /p-1 /s) > 0$ and a rbit r aryposi t iv e $ \ delta $. Be ca u se $p \geq d/2$, we ha ve a con tinuo us Sobolev em bedding $W ^ { 2 ,p}(X) \ subs e tL ^s(X)$ as alre ady o bserved, s o $$\|\xi \|_{L ^p(X)} \ leq C_1\d e l ta \|\xi \|_ {W_ {A_ 1}^ { 2 ,p }(X)} + \delt a ^ {-\n u} \|\xi\| _{L ^2(X)}, $$whe re$C_ 1= C_1(A_1 ,g,l,p)\i n[1 ,\ inf ty)$. Hence, f or $\ de lta (A_1, g ,l,p)\in ( 0,1] $gi v enby $\de l ta = 1/( 2C C_ 1)$, we c an us e re a rra ngement in to re pla c e $\ |\ xi \|_{L^p (X)}$ by $\|\ xi \|_{L^2(X) }$ . T herefo r e , the es timate implies $$\label { eq:W2p_ apr iori_ esti mate_Delt a_A _Sobol ev_ L 2_xi_r hs} \| \xi\| _{

8z) \in_(0, 1]$_in the preceding inequality,_rearrangement yields_the_desired estimate. Our_proof_of also verifies_that the operator_$\Delta_A$ in is bounded_since $\Delta_{A_s}$ is_bounded_with the same domain and range spaces. Next, we have the The finite-dimensionality of $\Ker\Delta_A\cap W_{A_1}^{2,p}(X;\Lambda^l\otimes\ad_P)$_follows from_Corollary_\[cor:Fredholmness\_and\_index\_Laplace\_operator\_on\_W2p\_Sobolev\_connection\]._We observe that, by increasing_the constant $C$ as needed,_the term_$\|\xi\|_{L^p(X)}$ appearing on the right-hand side of the_inequality_can be replaced_by $\|\xi\|_{L^2(X)}$. This is clear when $p \leq 2$,_while if $p > 2$, we_can choose $s_\in_(p,\infty)$_and apply the interpolation_inequality [@GilbargTrudinger Equation (7.10)], $$\|\xi\|_{L^p(X)} \leq_\delta \|\xi\|_{L^s(X)} + \delta^{-\nu}\|\xi\|_{L^2(X)},$$ for $\nu_:= (1/2-1/p)/(1/p-1/s) > 0$ and arbitrary positive_$\delta$. Because $p \geq d/2$, we_have a continuous Sobolev embedding_$W^{2,p}(X) \subset_L^s(X)$ as already observed, so_$$\|\xi\|_{L^p(X)} \leq C_1\delta_\|\xi\|_{W_{A_1}^{2,p}(X)} +_\delta^{-\nu}\|\xi\|_{L^2(X)},$$ where $C_1_= C_1(A_1,g,l,p) \in [1,\infty)$. Hence, for_$\delta(A_1,g,l,p) \in (0,1]$_given by $\delta = 1/(2CC_1)$, we_can_use rearrangement in_to_replace_$\|\xi\|_{L^p(X)}$ by_$\|\xi\|_{L^2(X)}$. Therefore, the_estimate_implies $$\label{eq:W2p_apriori_estimate_Delta_A_Sobolev_L2_xi_rhs} \|\xi\|_{

Hilbert space since $\mathcal{N}^\perp$ is closed in $\mathcal{H}^{2k}$. The restriction $S(\tau)|_{\mathcal{N}^\perp}$ defines a semigroup on $\mathcal{N}^\perp$ and its generator is the part of $L$ in $\mathcal{N}^\perp$, denoted by $L_{\mathcal{N}^\perp}$. The generator satisfies $$\begin{gathered} \mathrm{Re}(L_{\mathcal{N}^\perp}\mathbf{u}|\mathbf{u})_{\mathcal{N}^\perp}=\\ \mathrm{Re}\left ((L_0 D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H}+ (L'D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H} \right )-2k(D^{2k} \mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H} \\ \leq \left (\frac{1}{2}+pc_0-2k \right )\|\mathbf{u}\|_{\mathcal{N}^\perp}^2\end{gathered}$$ for all $\mathbf{u} \in \mathcal{D}(L_{\mathcal{N}^\perp})$ where we have used the commutator relation from Lemma \[lem\_elem\] and the fact that $L'$ and $D^{2k}$ commute, as already remarked. This, however, implies the growth estimate $\|S(\tau)\mathbf{u}\|_{\mathcal{N}^\perp}\leq e^{(\frac{1}{2}+pc_0-2k)\tau}\|\mathbf{u}\|_{\mathcal{N}^\perp}$ for all $\mathbf{u} \in \mathcal{N}^\perp$ and $\tau>0$ and by the equivalence of the norms $\|\cdot\|_{\mathcal{H}^{2k}}$ and $\|\cdot\|_{\mathcal{N}^\perp}$ (Proposition \[prop\_s0res\]) we arrive at the claim. Note that the growth estimate given in Proposition \[prop\_SonNperp\] is certainly not optimal, however, we do not make any attempts to improve it since we can make $k$ arbitrarily large. Analytic modes -------------- The point spectrum of $L$ can be calculated by solving a second order ODE. To this end we recall the definition of the operator $T(\lambda)$ from

Hilbert space since $ \mathcal{N}^\perp$ is closed in $ \mathcal{H}^{2k}$. The restriction $ S(\tau)|_{\mathcal{N}^\perp}$ defines a semigroup on $ \mathcal{N}^\perp$ and its generator is the character of $ L$ in $ \mathcal{N}^\perp$, announce by $ L_{\mathcal{N}^\perp}$. The generator satisfies $ $ \begin{gathered } \mathrm{Re}(L_{\mathcal{N}^\perp}\mathbf{u}|\mathbf{u})_{\mathcal{N}^\perp}=\\ \mathrm{Re}\left (( L_0 D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H}+ (L'D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H } \right) -2k(D^{2k } \mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H } \\ \leq \left (\frac{1}{2}+pc_0 - 2k \right) \|\mathbf{u}\|_{\mathcal{N}^\perp}^2\end{gathered}$$ for all $ \mathbf{u } \in \mathcal{D}(L_{\mathcal{N}^\perp})$ where we have used the commutator relation from Lemma \[lem\_elem\ ] and the fact that $ L'$ and $ D^{2k}$ commute, as already comment. This, however, implies the growth appraisal $ \|S(\tau)\mathbf{u}\|_{\mathcal{N}^\perp}\leq e^{(\frac{1}{2}+pc_0 - 2k)\tau}\|\mathbf{u}\|_{\mathcal{N}^\perp}$ for all $ \mathbf{u } \in \mathcal{N}^\perp$ and $ \tau>0 $ and by the equivalence of the norms $ \|\cdot\|_{\mathcal{H}^{2k}}$ and $ \|\cdot\|_{\mathcal{N}^\perp}$ (Proposition \[prop\_s0res\ ]) we arrive at the claim. notice that the increase estimate give in Proposition \[prop\_SonNperp\ ] is certainly not optimal, however, we do not seduce any attempts to improve it since we can make $ k$ arbitrarily large. Analytic modes -------------- The detail spectrum of $ L$ can be calculated by solving a second order ODE. To this conclusion we recall the definition of the operator $ T(\lambda)$ from

Hipbert space since $\mathcau{N}^\perp$ is closeb in $\mavhcal{H}^{2k}$. The resgriction $S(\tau)|_{\mathcal{N}^\perp}$ dehinew a stiigroup on $\mathcal{V}^\perp$ and its genwratie is the pedt of $L$ in $\mafmcal{N}^\'ecp$, denoted by $L_{\kathcal{N}^\pesp}$. The generatmr sctisfies $$\begin{gathered} \mathrm{Re}(L_{\mathcwl{N}^\perp}\kahhbf{u}|\mathbf{u})_{\majhcal{M}^\[erp}=\\ \jathrm{Re}\left ((L_0 D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mafhcal{H}+ (L'V^{2k}\mathbf{u}|D^{2k}\mathnf{u})_\mathcal{H} \right )-2k(D^{2k} \mathbv{u}|D^{2k}\lathbf{u})_\mathcal{H} \\ \lee \left (\frac{1}{2}+px_0-2k \rytht )\|\mathbf{u}\|_{\mxthcal{N}^\perk}^2\eud{gathered}$$ ror all $\mathbf{u} \in \mathcal{D}(L_{\matfcal{N}^\'erp})$ where qe hagg used the cimmutwtor relatiok from Nemma \[lrm\_elem\] and the fart tyat $L'$ and $D^{2k}$ commute, es already remarked. Jhis, howevar, implies the growrh estikate $\|S(\taj)\natfbf{h}\|_{\methdal{N}^\pegp}\lxq e^{(\frac{1}{2}+pc_0-2k)\tzu}\|\mathbf{u}\|_{\marhcal{N}^\perp}$ for all $\kaeybf{u} \in \mathcam{N}^\perp$ agd $\tau>0$ and by the equivalence of the nogms $\|\ddot\|_{\mathcal{H}^{2k}}$ and $\|\cdot\|_{\mqthcal{N}^\perp}$ (Propositiln \[prop\_s0rqs\]) we arrive at the claim. Note that the growth esthmate eivtn in Pfiplsition \[prop\_SonNperp\] is certainly not optimal, goeener, we do not makc any attempts to omorpde it since wg can mcie $k$ arbitrarily larhe. Analyjic moees -------------- The poynt xpectrum of $L$ can be calculqted by solvpng q second order ODE. To this enb we rgcall yhe definition of the o'eratod $T(\lambda)$ fgom

Hilbert space since $\mathcal{N}^\perp$ is closed in restriction defines a on $\mathcal{N}^\perp$ and of in $\mathcal{N}^\perp$, denoted $L_{\mathcal{N}^\perp}$. The generator $$\begin{gathered} \mathrm{Re}(L_{\mathcal{N}^\perp}\mathbf{u}|\mathbf{u})_{\mathcal{N}^\perp}=\\ \mathrm{Re}\left ((L_0 D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H}+ (L'D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H} )-2k(D^{2k} \mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H} \\ \leq \left (\frac{1}{2}+pc_0-2k \right )\|\mathbf{u}\|_{\mathcal{N}^\perp}^2\end{gathered}$$ for all $\mathbf{u} \in \mathcal{D}(L_{\mathcal{N}^\perp})$ we have used the commutator relation from Lemma \[lem\_elem\] and the fact that and commute, already This, however, implies the growth estimate $\|S(\tau)\mathbf{u}\|_{\mathcal{N}^\perp}\leq e^{(\frac{1}{2}+pc_0-2k)\tau}\|\mathbf{u}\|_{\mathcal{N}^\perp}$ for all $\mathbf{u} \in \mathcal{N}^\perp$ and $\tau>0$ and the equivalence of the norms $\|\cdot\|_{\mathcal{H}^{2k}}$ and $\|\cdot\|_{\mathcal{N}^\perp}$ \[prop\_s0res\]) we arrive at claim. Note that the growth given Proposition \[prop\_SonNperp\] certainly optimal, we do not any attempts to improve it since we can make $k$ arbitrarily large. Analytic modes -------------- The point of $L$ calculated by a order To this end the definition of the operator $T(\lambda)$

Hilbert space since $\mathcal{N}^\Perp$ is closEd in $\mAthCal{h}^{2k}$. the rEstrIction $S(\tau)|_{\mathCAl{N}^\pErp}$ defines a semigroup on $\MathcAl{n}^\Perp$ ANd Its geNerator IS tHE ParT oF $L$ In $\mAtHCaL{N}^\perP$, deNoted by $l_{\mathcal{N}^\pErp}$. thE generator saTIsFies $$\begin{gAthEred} \mathrm{Re}(l_{\maThcal{N}^\PeRp}\mAThbf{u}|\MatHbf{u})_{\mAthcal{n}^\Perp}=\\ \maThrm{Re}\lefT ((L_0 d^{2K}\mathbF{U}|D^{2k}\mathBF{U})_\mAthcAl{H}+ (L'D^{2k}\mathbf{u}|D^{2k}\maTHbF{U})_\mathcal{H} \right )-2K(D^{2k} \matHbF{U}|D^{2K}\MAthBf{u})_\Mathcal{H} \\ \leQ \lEft (\frAC{1}{2}+pc_0-2k \rigHT )\|\mATHBf{u}\|_{\MAthcal{N}^\perp}^2\enD{gathered}$$ foR All $\Mathbf{U} \iN \maTHcal{D}(L_{\MathcAl{n}^\PerP})$ where we havE useD the commuTator rELation fROm Lemma \[Lem\_eleM\] anD thE facT ThAt $l'$ anD $D^{2K}$ ComMUtE, as ALreAdy remarKeD. THis, hoWeveR, IMPLies The GrowTh estImate $\|S(\tau)\mathBf{u}\|_{\MathCAl{N}^\Perp}\lEq e^{(\frAc{1}{2}+pc_0-2K)\tAu}\|\matHbf{u}\|_{\maThcal{n}^\pErp}$ for all $\mathbf{U} \in \mAthcal{N}^\peRp$ aNd $\Tau>0$ AnD by thE EquivaLenCe oF the norMs $\|\cdot\|_{\mAThcAl{h}^{2K}}$ ANd $\|\Cdot\|_{\mathcal{N}^\perp}$ (PrOpOSItIon \[prop\_s0Res\]) we aRRiVe AT the claiM. NOte That THE growTh esTImAte given In PropOSiTiOn \[prop\_SOnnperp\] iS cErtAinLy not OPtimAl, howeVer, we do nOt makE Any attempts to iMProve it since wE CaN MAkE $K$ arbItrArily large. ANalyTIc moDes -------------- THE pOinT SpectRum of $l$ cAN bE Calculated by solving A sEcond oRder OdE. To this end we Recall the dEFINition of The oPErATor $T(\lambda)$ from

Hilbert space since $\mat hcal{N}^\p erp$isclo se d in $\m athcal{H}^{2k} $ . Th e restriction $S(\tau) |_{\m at h cal{ N }^ \perp }$ defi n es a se mi gr oup o n $ \math cal {N}^\pe rp$ and it s g en erator is th e p art of $L$ in $\mathcal{N }^\ perp$, d eno t ed by $L _{\ma thcal{ N }^\per p}$. Thege n erator satisfi e s $ $\be gin{gathered} \ma t hr m {Re}(L_{\mathc al{N}^ \p e rp } \ mat hbf {u}|\mathb f{ u})_{ \ mathcal { N} ^ \ p erp } =\\ \mathrm{R e}\left ((L _ 0 D ^{2k}\ ma thb f {u}|D^ {2k}\ ma t hbf {u})_\mathc al{H }+ (L'D^{ 2k}\ma t hbf{u}| D ^{2k}\m athbf{ u}) _\m athc a l{ H} \r ig h t ) - 2k (D^ { 2k} \mathbf {u }| D^{2k }\ma t h b f {u}) _\m athc al{H} \\ \leq \lef t ( \fra c {1} {2}+p c_0-2 k \r ig ht )\ |\math bf{u} \| _{\mathcal{N}^\ perp }^2\end{g ath er ed} $$ fora ll $\m ath bf{ u} \in\mathca l {D} (L _ { \ ma thcal{N}^\perp})$wh e r ewe haveused t h eco m mutatorre lat ionf r om Le mma\ [l em\_elem \] and th efact th at $L'$an d $ D^{ 2k}$c ommu te, as already rema r ked. This, how e ver, impliest he g ro w th e sti mate $\|S(\ tau) \ math bf{u } \| _{\ m athca l{N}^ \p e rp } \leq e^{(\frac{1}{2 }+ pc_0-2 k)\ta u}\|\mathbf{u }\|_{\math c a l {N}^\per p}$f or all $\mathbf{u } \in \mathcal{ N }^\perp$ and$\tau>0$ and by t h e equival enc e o f t hen o rm s $\|\cdot\|_ { \ math ca l{H}^{2 k}} $ and $ \|\ cdo t\| _{\ ma thcal{N}^ \perp}$(P ro po si tio n \[p r op\_s0re s\ ])we ar rivea t theclaim . N ot et hat the gr o wt h esti ma te giv enin Prop osit i on\[prop\ _SonNperp \]i s ce rt ai nly not optimal, how ev er, we dono t m ake an y attempts to improve it since we can mak e $ k$ ar bitr arily lar ge. Anal yti c modes ----- ----- -- --T he po i n tspe ct rum of $L$ c anbe ca lc ulat ed by s olving a second or d erODE. To thisend wer e ca llt he def in i tio n of the operator $T(\lambd a) $ f rom

Hilbert_space since_$\mathcal{N}^\perp$ is closed in_$\mathcal{H}^{2k}$. The_restriction_$S(\tau)|_{\mathcal{N}^\perp}$ defines_a_semigroup on $\mathcal{N}^\perp$_and its generator_is the part of_$L$ in $\mathcal{N}^\perp$,_denoted_by $L_{\mathcal{N}^\perp}$. The generator satisfies $$\begin{gathered} \mathrm{Re}(L_{\mathcal{N}^\perp}\mathbf{u}|\mathbf{u})_{\mathcal{N}^\perp}=\\ \mathrm{Re}\left ((L_0 D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H}+ (L'D^{2k}\mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H} \right )-2k(D^{2k} \mathbf{u}|D^{2k}\mathbf{u})_\mathcal{H} \\ \leq \left (\frac{1}{2}+pc_0-2k \right )\|\mathbf{u}\|_{\mathcal{N}^\perp}^2\end{gathered}$$ for_all_$\mathbf{u} \in_\mathcal{D}(L_{\mathcal{N}^\perp})$_where_we have used the commutator_relation from Lemma \[lem\_elem\] and_the fact_that $L'$ and $D^{2k}$ commute, as already remarked._This,_however, implies the_growth estimate $\|S(\tau)\mathbf{u}\|_{\mathcal{N}^\perp}\leq e^{(\frac{1}{2}+pc_0-2k)\tau}\|\mathbf{u}\|_{\mathcal{N}^\perp}$ for all $\mathbf{u} \in \mathcal{N}^\perp$ and $\tau>0$_and by the equivalence of the_norms $\|\cdot\|_{\mathcal{H}^{2k}}$ and_$\|\cdot\|_{\mathcal{N}^\perp}$_(Proposition_\[prop\_s0res\]) we arrive at_the claim. Note that the growth estimate_given in Proposition \[prop\_SonNperp\] is certainly_not optimal, however, we do not make_any attempts to improve it since_we can make $k$ arbitrarily_large. Analytic modes -------------- The_point spectrum of $L$ can_be calculated by_solving a_second order ODE._To this end we recall the_definition of the_operator $T(\lambda)$ from

$\rho_0$. (a)-(f) correspond to $\rho_0=0.06$, $0.08$, $0.10$, $0.12$, $0.14$ and $0.16$. []{data-label="fig_PFC_vac"}](fig3f "fig:"){width="0.33\columnwidth"} ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- With the vacancy term, Eq. (\[eqn\_vac\_f\]), we can numerically verify the analytical calculation for the coexistence between the periodic phase and vacancies. Fig. (\[fig\_PFC\_vac\]) provides results from simulations with $r=-0.9$ and different values of $\rho_0$, showing clearly that the number of atoms increases with $\rho_0$. In addition, Fig. (\[fig\_pfc\_density\]) shows that the PFC atomic density (*i.e.,* the number of atoms per unit area.) indeed increases linearly with $\rho_0$, as expected. The curve starts to saturate at around $\rho_0=0.15$, as opposed to the prediction from Eq (\[eqn\_vac\_phase\]), $\rho_{min} \approx 0.134$. This discrepancy is not surprising for several reasons: In the analytical calculation, we consider only the one-mode approximation in the ansatz; we did not account for the surface energy between the triangular phase and the vacancies; and we did not account for thermal fluctuations, introduced in the simulation to help the system equilibrate faster. ![ The PFC atomic density increases linearly with the order parameter, $\rho_0$, when the vacancy term is added to the model. $r=-0.9$ is used. The curve starts to saturate at around $\rho_0=0.15$, as opposed to the analytical prediction $\rho_0=\sqrt{(-48-56r)/133}=0.134$. []{data-label="fig_pfc_density"}](fig4){width="0.75\columnwidth"} With the modifications described above,the PFC simulation operates very much like a molecular dynamics simulation, but on diffusive time scales many orders of magnitude faster than pure molecular dynamics[@Elder04]. We can control the number of atoms and the temperature in the system by adjusting $\rho_0$ and the magnitude of thermal noise, $\eta$, respectively. The interaction potential between individual PFC atoms is specified by the PFC free energy

$ \rho_0$. (a)-(f) correspond to $ \rho_0=0.06 $, $ 0.08 $, $ 0.10 $, $ 0.12 $, $ 0.14 $ and $ 0.16$. [ ] { data - label="fig_PFC_vac"}](fig3f " fig:"){width="0.33\columnwidth " } ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- With the vacancy term, Eq.   (\[eqn\_vac\_f\ ]), we can numerically control the analytic calculation for the coexistence between the periodic phase and vacancies. Fig.   (\[fig\_PFC\_vac\ ]) provides results from simulation with $ r=-0.9 $ and different values of $ \rho_0 $, show distinctly that the number of atom increases with $ \rho_0$. In addition, Fig.   (\[fig\_pfc\_density\ ]) show that the PFC atomic density (* i.e., * the phone number of atom per unit area .) indeed increases linearly with $ \rho_0 $, as expected. The curve start to saturate at around $ \rho_0=0.15 $, as opposed to the prediction from Eq   (\[eqn\_vac\_phase\ ]), $ \rho_{min } \approx 0.134$. This discrepancy is not surprising for several rationality: In the analytical calculation, we consider only the one - mode estimate in the ansatz; we did not account for the airfoil energy between the triangular phase and the vacancies; and we did not account for thermal fluctuation, introduced in the simulation to help the system equilibrate faster. ! [ The PFC atomic density increases linearly with the order parameter, $ \rho_0 $, when the vacancy term is added to the model. $ r=-0.9 $ is used. The curve start to saturate at around $ \rho_0=0.15 $, as opposed to the analytic prediction $ \rho_0=\sqrt{(-48 - 56r)/133}=0.134$. [ ] { datum - label="fig_pfc_density"}](fig4){width="0.75\columnwidth " } With the change described above, the PFC simulation operate on very much like a molecular dynamics simulation, but on diffusive meter scales many orders of magnitude faster than pure molecular dynamics[@Elder04 ]. We can control the phone number of atoms and the temperature in the system by adjusting $ \rho_0 $ and the magnitude of thermal randomness, $ \eta$, respectively. The interaction potential between individual PFC atom is assign by the PFC free energy

$\rhl_0$. (a)-(f) correspond to $\rho_0=0.06$, $0.08$, $0.10$, $0.12$, $0.14$ and $0.16$. []{data-labgl="dig_PFC_tac"}](fig3f "fig:"){widtf="0.33\columnwidth"} ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- With the vacanry twrm, Ew. (\[eqn\_vac\_f\]), we can numerizally verpfy the abalyuical calculation for the coexiabence uetween the perlodic phase and vacancies. Fkg. (\[yig\_PFC\_vac\]) provides results from simujations wlth $r=-0.9$ and diffgrent daluss of $\rho_0$, showing clearly that the number of atoms incteases with $\rho_0$. In additioj, Fih. (\[fig\_pfc\_density\]) shoas that the PFC qtomic densigy (*i.e.,* the number of atkms per unit area.) indeed increares lnnearly wity $\eho_0$, ds expected. The burve starts bp satusate at around $\rho_0=0.15$, as op'osee to the prediction fcom Eq (\[eqn\_vac\_phase\]), $\rhj_{min} \apprmx 0.134$. This discrepanct us noj sur[riskbg wor sxvedal rewsois: In the ahalytical cqlculation, we consicew only the one-jode a[pwoximation in the ansatz; we did not accmunf for the surface energt between the triangupar phase and the vacancies; and we did not account for thesmal hljctbqtionr, ijtroduced in the simulation to help the systei eauplibrate faster. ![ Tme PFC atomic densotj ogcreases linexrly wnfh the order parametfr, $\rho_0$, rhen rhe vacansy trrm is added to the model. $r=-0.9$ is used. The xurve starts to sacurate at aruund $\rho_0=0.15$, ss opposed to the analycical lrediction $\gho_0=\sqrt{(-48-56r)/133}=0.134$. []{sxta-label="fig_pfc_devsiny"}](fic4){width="0.75\columnwidth"} With the iodificatmons bescribea abpve,the PFC simulwtion operates very much like a molecular dynamics simulation, but on difhnsive time scslas kany ordzrs of magnitude fafter than pure molecujar dhnamics[@Eldvr04]. We can control thq number of adlms and the vemperatuwe ib thw systeo by adjusting $\tho_0$ and tkt magnitudw of thermal noise, $\etx$, respectively. Tkt ibteraction potemtixl fenwexn inqhvidual PFC dtomr ir speckfied by thc PWC ftee energy

$\rho_0$. (a)-(f) correspond to $\rho_0=0.06$, $0.08$, $0.10$, and []{data-label="fig_PFC_vac"}](fig3f "fig:"){width="0.33\columnwidth"} With the vacancy numerically the analytical calculation the coexistence between periodic phase and vacancies. Fig. (\[fig\_PFC\_vac\]) results from simulations with $r=-0.9$ and different values of $\rho_0$, showing clearly that number of atoms increases with $\rho_0$. In addition, Fig. (\[fig\_pfc\_density\]) shows that the atomic (*i.e.,* number atoms per unit area.) indeed increases linearly with $\rho_0$, as expected. The curve starts to saturate around $\rho_0=0.15$, as opposed to the prediction from (\[eqn\_vac\_phase\]), $\rho_{min} \approx 0.134$. discrepancy is not surprising for reasons: the analytical we only one-mode approximation in ansatz; we did not account for the surface energy between the triangular phase and the vacancies; and did not thermal fluctuations, in simulation help the system ![ The PFC atomic density increases order parameter, $\rho_0$, when the vacancy term is to the $r=-0.9$ is used. The curve starts saturate at around $\rho_0=0.15$, as opposed to the prediction $\rho_0=\sqrt{(-48-56r)/133}=0.134$. []{data-label="fig_pfc_density"}](fig4){width="0.75\columnwidth"} With the modifications described above,the PFC simulation operates very much like a simulation, but on diffusive scales many orders magnitude than molecular We can the number of atoms and the temperature in the system by $\rho_0$ and the magnitude of thermal noise, $\eta$, respectively. The between PFC atoms is by the PFC free

$\rho_0$. (a)-(f) correspond to $\rho_0=0.06$, $0.08$, $0.10$, $0.12$, $0.14$ and $0.16$. []{daTa-label="fig_pFC_vaC"}](fiG3f "fIg:"){WidtH="0.33\colUmnwidth"} ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- With thE VacaNcy term, Eq. (\[eqn\_vac\_f\]), we can nUmeriCaLLy veRIfY the aNalyticAL cALCulAtIoN foR tHE cOexisTenCe betweEn the perioDic PhAse and vacancIEs. fig. (\[fig\_PFC\_vAc\]) pRovides resulTs fRom simUlAtiONs witH $r=-0.9$ aNd difFerent VAlues oF $\rho_0$, showiNg CLearly THat the nUMBeR of aToms increases with $\RHo_0$. iN addition, Fig. (\[fiG\_pfc\_deNsITy\]) SHOws ThaT the PFC atoMiC densITy (*i.e.,* the NUmBER Of aTOms per unit areA.) indeed incrEAseS lineaRlY wiTH $\rho_0$, as ExpecTeD. the Curve starts To saTurate at aRound $\rHO_0=0.15$, as oppoSEd to the PredicTioN frOm Eq (\[EQn\_VaC\_phAsE\]), $\Rho_{MIn} \AppROx 0.134$. THis discrEpAnCy is nOt suRPRISing For SeveRal reAsons: In the anaLytIcal CAlcUlatiOn, we cOnsiDeR only The one-Mode aPpRoximation in the AnsaTz; we did noT acCoUnt FoR the sURface eNerGy bEtween tHe trianGUlaR pHASE aNd the vacancies; and wE dID NoT account For theRMaL fLUctuatioNs, IntRoduCED in thE simULaTion to heLp the sYStEm EquilibRaTe fastEr. ![ the pFC AtomiC DensIty incReases liNearlY With the order paRAmeter, $\rho_0$, when THe VACaNCy teRm iS added to the ModeL. $R=-0.9$ is uSed. THE cUrvE StartS to saTuRAtE At around $\rho_0=0.15$, as opposeD tO the anAlytiCal prediction $\Rho_0=\sqrt{(-48-56r)/133}=0.134$. []{daTA-LAbel="fig_pFc_deNSiTY"}](fig4){width="0.75\columNwidtH"} With the moDIficatioNs desCribed abOve,the PFC SIMulation OpeRatEs vEry MUCh Like a moleculaR DYnamIcS simulaTioN, but on dIffUsiVe tIme ScAles many oRders of mAgNiTuDe FasTer thAN pure molEcUlaR dYnaMics[@ELDer04]. We cAn conTrol ThE nUMbeR of atomS AnD THe teMpErAturE in ThE systEm by ADjuSting $\rhO_0$ and the maGniTUde oF tHeRmal noiSe, $\eta$, respectiVeLy. The interAcTioN potenTIAl betweeN individual PFC atoms is spECified bY thE PFC fRee eNergy

$\rho_0$. (a)-(f) corresp ond to $\r ho_0= 0.0 6$, $ 0.08 $, $ 0.10$, $0.12$, $0.1 4$ and $0.16$. []{data -labe l= " fig_ P FC _vac" }](fig3 f " f i g:" ){ wi dth =" 0 .3 3\col umn width"} ------- --- -- ------------ - -- ---------- --- ------------ --- ------ -- --- - ----- --- ----- ------ - ------ --------- -- - ------ - ------- - - -- ---- ----------------- - -- - -------------- ------ -- - -- - - --- --- ---------- -- ----- - ------- - -- - - - --- - ------------- ----------- - --- ------ -- --- - ------ ----- -- - --- ----------- ---- --------- ------ - ------- - ------- ------ --- --- ---- - -- -- --- -- - --- - -- --- - --- -------- -- -- ----- ---- - - - - ---- --- ---- ----- ------------- --- ---- - --- ----- ----- ---- -- ----- ------ ----- -- --------------- ---- --------- --- -- --- -- ----- - ------ --- --- ------- ------- - --- -- - - - -- ------------------ -- - - -- -------- ------ - -- -- - -------- -- --- ---- - - ----- ---- - -- -------- ------ - -- -- ------- -- ------ -- --- --- ----- - ---- ------ -------- ----- - -------------- - ------------- - -- - - -- - ---- --- ----------- ---- - ---- ---- - -- --- - ----- ----- -- - -- - ----------- With t he vacan cy te rm, Eq. (\[eq n\_vac\_f\ ] ) , we cannume r ic a lly verify the anal ytical cal c ulationfor t he coexi stence be t w een theper iod icpha s e a nd vacancies. F ig.(\ [fig\_P FC\ _vac\]) pr ovi des re su lts fromsimulati on swi th $r =-0.9 $ and dif fe ren tval ues o f $\rho _0$,show in gc lea rly tha t t h e num be rof a tom sincre ases wit h $\rho _0$. In a ddi t ion, F ig . (\[fi g\_pfc\_densi ty \]) showsth atthe PF C atomic d ensity (*i.e.,* the num b er of a tom s per uni t area.)ind eed in cre a ses li nearly with $ \rh o _ 0$, a s ex pec te d. The cur v e st artsto sat urate a t around $\rho_0=0 . 15$ , as opposedtothep r ed ict i on fro mE q ( \ [ eqn\_vac\_phase \]), $\rho _{ m in } \approx0 .13 4$ . Thisdiscrep ancyi s not s urprising for seve ra l re a s ons : In the a nalytica l calcula t ion,w econsi der onlyth e o ne-mo de app r oxi matio n in t he ansat z; we d id not a ccount for the surfaceenergy betw een the tria ngu l arphase and the vacancies ; a ndwe di d n o t acc ount fo r t h ermal flu c tuations, in tro d u ce d in the si m u l ati on to he l p thesyst em equilibrate fa s ter. ![ The P FC a t o mic de n sity i ncreases linea rly w i t h the or de r parameter , $\rho_ 0$ , when the v acancy term i s ad d ed tothemod el. $r=-0 .9$ i s used.Th ec urve s tart sto sat uratea t ar o u nd $\rho_0=0.15$ , aso p posed tothe a na lytical pred iction $\r ho_0=\sqrt{ (-48-5 6r)/ 133}= 0.134$. [ ]{data -la be l="fig_pfc _ density"} ](fig 4){widt h= "0.7 5\c olumnw idth " } Wit h th emod ification s de s cr ib e d a bove ,thePF C si mulationo peratesver y much l ik e a m olecul a rd y namics sim ula tion, b ut on diff u sive ti m e sca les ma ny ord ers ofm agn it ude fas ter t han puremolecular dyn am ics[ @Elder04 ]. Weca n c on trol the number of atom s and th e temper atu r e in the sy st em by adjust ing $ \r h o_0$ an d the magnit u deofthe rmal no ise, $\eta$, respectively. Th e in terac t ion pote ntial b etween i n divi dual PFC atoms is spec ified by the PF C fr e e en erg y

$\rho_0$._(a)-(f) correspond_to $\rho_0=0.06$, $0.08$, $0.10$,_$0.12$, $0.14$_and_$0.16$. []{data-label="fig_PFC_vac"}](fig3f_"fig:"){width="0.33\columnwidth"} _ ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- With the_vacancy term, Eq. (\[eqn\_vac\_f\]),_we can numerically verify_the analytical calculation_for_the coexistence between the periodic phase and vacancies. Fig. (\[fig\_PFC\_vac\]) provides results from simulations with_$r=-0.9$_and different_values_of_$\rho_0$, showing clearly that the_number of atoms increases with_$\rho_0$. In_addition, Fig. (\[fig\_pfc\_density\]) shows that the PFC atomic density_(*i.e.,*_the number of_atoms per unit area.) indeed increases linearly with $\rho_0$,_as expected. The curve starts to_saturate at around_$\rho_0=0.15$,_as_opposed to the prediction_from Eq (\[eqn\_vac\_phase\]), $\rho_{min} \approx 0.134$. This_discrepancy is not surprising for several_reasons: In the analytical calculation, we consider_only the one-mode approximation in the_ansatz; we did not account_for the_surface energy between the triangular_phase and the_vacancies; and_we did not_account for thermal fluctuations, introduced in_the simulation to_help the system equilibrate faster. ![ The_PFC_atomic density increases_linearly_with_the order_parameter, $\rho_0$, when_the_vacancy term_is_added to the model. $r=-0.9$ is_used._The curve starts to saturate at around_$\rho_0=0.15$, as opposed to_the_analytical prediction $\rho_0=\sqrt{(-48-56r)/133}=0.134$. []{data-label="fig_pfc_density"}](fig4){width="0.75\columnwidth"} With_the modifications described above,the PFC_simulation operates very much like a_molecular dynamics_simulation, but_on diffusive time scales many orders of magnitude faster than pure_molecular dynamics[@Elder04]. We can control the_number of atoms and_the temperature_in_the system by_adjusting_$\rho_0$ and_the magnitude of thermal noise, $\eta$, respectively._The interaction_potential between individual PFC atoms is_specified by the PFC_free_energy

on $M$ to a vector field $\mathbf{u}^{\#}$ on the tangent bundle $TM$. The components of $\mathbf{u}^{\#}$ with respect to the coordinate system $(x^i,y^i)$ are $$\begin{aligned} \left(u^i(x),\frac{\partial u^i}{\partial x^j}(x) y^j\right). \end{aligned}$$ Now the associated differential operator for $\mathbf{u}^{\#}\in\mathrm{Vect}(TM)$ that acts on smooth functions on $TM$ will be $$u^i(x)\frac{\partial}{\partial x^i} + \frac{\partial u^i}{\partial x^j}(x)y^j \frac{\partial}{\partial y^i}.$$ The crucial property of complete lifts is that the assignment $\mathbf{u}\mapsto\mathbf{u}^{\#}$ is a *Lie algebra homomorphism* $\mathrm{Vect}(M)\rightarrow\mathrm{Vect}(TM)$ with respect to the commutator bracket of vector fields on $M$ and $TM$ respectively. Symbolically this is $$\begin{aligned} \label{homo1} \left([\mathbf{u},\mathbf{v}]_M\right)^{\#} = \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM},\end{aligned}$$ which can be checked by direct computation or otherwise. (The subscripts indicate the space on which the vector fields live.) A proof sketch can be found in appendix **\[app:diff\]**. This immediately implies that vector fields $\mathbf{u}\in\mathrm{Vect}(M)$ on $M$ can act on functions $f\in C^\infty(TM)$ on $TM$ via $f\mapsto \mathbf{u}^{\#}\cdot f$, and moreover this action is a Lie algebra representation: $$\begin{aligned} \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM} \cdot f = \mathbf{u}^{\#}\cdot\mathbf{v}^{\#}\cdot f - \mathbf{v}^{\#}\cdot\mathbf{u}^{\#}\cdot f = \left([\mathbf{u},\mathbf{v}]_M\right)^{\#} \cdot f.\end{aligned}$$ Thus we can add this representation of $\mathrm{Vect}(M)$ on $

on $ M$ to a vector field $ \mathbf{u}^{\#}$ on the tangent bundle $ TM$. The components of $ \mathbf{u}^{\#}$ with respect to the coordinate arrangement $ (x^i, y^i)$ are $ $ \begin{aligned } \left(u^i(x),\frac{\partial u^i}{\partial x^j}(x) y^j\right). \end{aligned}$$ nowadays the associated differential operator for $ \mathbf{u}^{\#}\in\mathrm{Vect}(TM)$ that acts on fluent functions on $ TM$ will be $ $ u^i(x)\frac{\partial}{\partial x^i } + \frac{\partial u^i}{\partial x^j}(x)y^j \frac{\partial}{\partial y^i}.$$ The crucial place of complete lifts is that the assignment $ \mathbf{u}\mapsto\mathbf{u}^{\#}$ is a * Lie algebra homomorphism * $ \mathrm{Vect}(M)\rightarrow\mathrm{Vect}(TM)$ with esteem to the commutator bracket of vector fields on $ M$ and $ TM$ respectively. Symbolically this is $ $ \begin{aligned } \label{homo1 } \left([\mathbf{u},\mathbf{v}]_M\right)^{\ # } = \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM},\end{aligned}$$ which can be checked by lineal computation or otherwise. (The subscripts indicate the quad on which the vector fields live .) A proof sketch can be found in appendix * * \[app: diff\ ] * *. This immediately imply that vector fields $ \mathbf{u}\in\mathrm{Vect}(M)$ on $ M$ can act on functions $ f\in C^\infty(TM)$ on $ TM$ via $ f\mapsto \mathbf{u}^{\#}\cdot f$, and furthermore this action is a Lie algebra representation: $ $ \begin{aligned } \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM } \cdot f = \mathbf{u}^{\#}\cdot\mathbf{v}^{\#}\cdot f - \mathbf{v}^{\#}\cdot\mathbf{u}^{\#}\cdot f = \left([\mathbf{u},\mathbf{v}]_M\right)^{\ # } \cdot f.\end{aligned}$$ therefore we can add this theatrical performance of $ \mathrm{Vect}(M)$ on $

on $M$ to a vector field $\matmbf{u}^{\#}$ on the tanggnr bundne $TM$. Fhe compunents of $\mathbf{u}^{\#}$ with respert ti the coordinate system $(x^i,y^k)$ are $$\begpn{aligned} \lefu(u^i(x),\frac{\partial u^m}{\lartial x^j}(x) y^n\vight). \xnd{aligned}$$ Now jhe associatad differentian upzrator for $\mathbf{u}^{\#}\in\mathrm{Vect}(TM)$ thae acts pn smooth functijns pg $TM$ will be $$u^i(x)\frac{\partial}{\partial x^i} + \frac{\pagtial u^i}{\partial x^k}(x)y^j \frac{\partial}{\partial y^i}.$$ The crucial property lf complete lifew is that thd assignment $\mathbf{u}\maksto\mathbf{u}^{\#}$ is a *Lie algebra homumorpkism* $\mathrm{Cext}(M)\ghghtarrow\mavhrm{Vebt}(TM)$ with resivct to dhe comkutator brackeb of tectir fields on $M$ and $TM$ respectively. Symbjlically dhns is $$\begin{aligned} \labwl{yomo1} \laft([\mxrhbw{u},\mztibf{b}]_M\righh)^{\#} = \meft[\mathbf{h}^{\#},\mathbf{v}^{\#}\rigyt]_{TM},\end{aligned}$$ whicn sqn be checked by diwest computation or otherwise. (The subscrikts ihdicate the space on whuch the vector fields live.) A pwoof sketch can be found in appendix **\[app:diff\]**. This hmmedmagelv implkws that vector fields $\mathbf{u}\in\mathrm{Vect}(M)$ on $M$ dam sct on functioks $f\in C^\infty(TM)$ on $TL$ fya $f\mapsto \majhbf{u}^{\#}\cdot r$, and moreover thid actiog is q Lie algtbra tepresentation: $$\begin{aligned} \oeft[\mathbf{u}^{\#},\mcthvf{v}^{\#}\right]_{TM} \cdot f = \mathbf{u}^{\#}\cdoc\mathbg{v}^{\#}\cdoy f - \mathbf{v}^{\#}\cdot\mathbf{u}^{\#}\edot f = \left([\mathbv{u},\mathbf{v}]_J\fight)^{\#} \cdot f.\end{auigked}$$ Thus we can add this reprqsentatioi of $\kathrm{Vdct}(M)$ on $

on $M$ to a vector field $\mathbf{u}^{\#}$ tangent $TM$. The of $\mathbf{u}^{\#}$ with $(x^i,y^i)$ $$\begin{aligned} \left(u^i(x),\frac{\partial u^i}{\partial y^j\right). \end{aligned}$$ Now associated differential operator for $\mathbf{u}^{\#}\in\mathrm{Vect}(TM)$ that on smooth functions on $TM$ will be $$u^i(x)\frac{\partial}{\partial x^i} + \frac{\partial u^i}{\partial x^j}(x)y^j y^i}.$$ The crucial property of complete lifts is that the assignment $\mathbf{u}\mapsto\mathbf{u}^{\#}$ is *Lie homomorphism* with to the commutator bracket of vector fields on $M$ and $TM$ respectively. Symbolically this is $$\begin{aligned} \left([\mathbf{u},\mathbf{v}]_M\right)^{\#} = \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM},\end{aligned}$$ which can be checked by computation or otherwise. (The indicate the space on which vector live.) A sketch be in appendix **\[app:diff\]**. immediately implies that vector fields $\mathbf{u}\in\mathrm{Vect}(M)$ on $M$ can act on functions $f\in C^\infty(TM)$ on $TM$ via \mathbf{u}^{\#}\cdot f$, this action a algebra $$\begin{aligned} \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM} \cdot \mathbf{u}^{\#}\cdot\mathbf{v}^{\#}\cdot f - \mathbf{v}^{\#}\cdot\mathbf{u}^{\#}\cdot f = Thus we can add this representation of $\mathrm{Vect}(M)$ $

on $M$ to a vector field $\mathbf{u}^{\#}$ oN the tangenT bundLe $Tm$. ThE cOmpoNentS of $\mathbf{u}^{\#}$ with REspeCt to the coordinate systeM $(x^i,y^i)$ ArE $$\BegiN{AlIgned} \Left(u^i(x),\FRaC{\PArtIaL u^I}{\paRtIAl X^j}(x) y^j\RigHt). \end{alIgned}$$ Now thE asSoCiated differENtIal operatoR foR $\mathbf{u}^{\#}\in\maThrM{Vect}(Tm)$ tHat ACts on SmoOth fuNctionS On $TM$ wiLl be $$u^i(x)\frAc{\PArtial}{\PArtial x^I} + \FRaC{\parTial u^i}{\partial x^j}(x)y^J \FrAC{\partial}{\partiaL y^i}.$$ The CrUCiAL ProPerTy of compleTe Lifts IS that thE AsSIGNmeNT $\mathbf{u}\mapstO\mathbf{u}^{\#}$ is a *lIe aLgebra HoMomORphism* $\MathrM{VECt}(M)\Rightarrow\mAthrM{Vect}(TM)$ wiTh respECt to the COmmutatOr bracKet Of vEctoR FiElDs oN $M$ ANd $Tm$ ReSpeCTivEly. SymboLiCaLly thIs is $$\BEGIN{aliGneD} \labEl{homO1} \left([\mathbf{u},\mAthBf{v}]_M\RIghT)^{\#} = \left[\MathbF{u}^{\#},\maThBf{v}^{\#}\riGht]_{TM},\eNd{aliGnEd}$$ which can be cheCked By direct cOmpUtAtiOn Or othERwise. (THe sUbsCripts iNdicate THe sPaCE ON wHich the vector fieldS lIVE.) A Proof skeTch can BE fOuND in appenDiX **\[apP:difF\]**. tHis imMediATeLy implieS that vECtOr Fields $\mAtHbf{u}\in\MaThrM{VeCt}(M)$ on $m$ Can aCt on fuNctions $f\In C^\inFTy(TM)$ on $TM$ via $f\maPSto \mathbf{u}^{\#}\cdoT F$, aND MoREoveR thIs action is a lie aLGebrA repREsEntATion: $$\bEgin{aLiGNeD} \Left[\mathbf{u}^{\#},\mathbf{v}^{\#}\rIgHt]_{TM} \cdOt f = \maThbf{u}^{\#}\cdot\mathBf{v}^{\#}\cdot f - \maTHBF{v}^{\#}\cdot\maThbf{U}^{\#}\CdOT f = \left([\mathbf{u},\mAthbf{V}]_M\right)^{\#} \cdoT F.\end{aligNed}$$ ThUs we can aDd this repRESentatioN of $\MatHrm{vecT}(m)$ On $

on $M$ to a vector field$\mathbf{u }^{\# }$onth e ta ngen t bundle $TM$. Thecomponents of $\mathbf {u}^{ \# } $ wi t hrespe ct to t h ec o ord in at e s ys t em $(x^ i,y ^i)$ ar e $$\begin {al ig ned} \left( u ^i (x),\frac{ \pa rtial u^i}{\ par tial x ^j }(x ) y^j\ rig ht).\end{a l igned} $$ Now th ea ssocia t ed diff e r en tial operator for $\m a th b f{u}^{\#}\in\m athrm{ Ve c t} ( T M)$ th at acts on s mooth functio n so n $TM $ will be $$u^ i(x)\frac{\ p art ial}{\ pa rti a l x^i} + \f ra c {\p artial u^i} {\pa rtial x^j }(x)y^ j \frac{ \ partial }{\par tia l y ^i}. $ $Th e c ru c ial pr ope r tyof compl et elifts ist h a t the as sign ment$\mathbf{u}\m aps to\m a thb f{u}^ {\#}$ isa*Liealgebr a hom om orphism* $\math rm{V ect}(M)\r igh ta rro w\ mathr m {Vect} (TM )$with re spect t o th ec o m mu tator bracket of v ec t o rfields o n $M$a nd $ T M$ respe ct ive ly.S y mboli call y t his is $ $\begi n {a li gned} \ la bel{ho mo 1} \lef t ([\m athbf{ u},\math bf{v} ] _M\right)^{\#} = \left[\math b f{ u } ^{ \ #},\ mat hbf{v}^{\#} \rig h t]_{ TM}, \ en d{a l igned }$$ w hi c hc an be checked by di re ct com putat ion or otherw ise. (Thes u b scriptsindi c at e the space onwhich the vecto r fieldslive. ) A proo f sketchc a n be fou ndinapp end i x * *\[app:diff\] * * . T hi s immed iat ely imp lie s t hat ve ct or fields $\mathb f{ u} \i n\ mat hrm{V e ct}(M)$on $M $can acto n func tions $f\ in C ^ \in fty(TM) $ o n $TM$ v ia $f\ map st o \ma thbf { u}^ {\#}\cd ot f$, an d m o reov er t his act ion is a Lieal gebra repr es ent ation: $ $\begin{ aligned} \left[\mathbf{ u }^{\#}, \ma thbf{ v}^{ \#}\right ]_{ TM} \c dot f = \m athbf{ u}^{\ #} \cd o t \math b f {v }^{ \# }\cdot f - \ mat hbf{v }^ {\#} \cdot\m athbf{u}^{\#}\cdot f = \left([\math bf{ u},\ m a th bf{ v }] _ M\r ig h t)^ { \ #} \cdot f.\end {aligned}$ $T hu s we can a d d t hi s repre sentati on of $\mathr m{Vect}(M )$ on $

on_$M$ to_a vector field $\mathbf{u}^{\#}$_on the_tangent_bundle $TM$._The_components of $\mathbf{u}^{\#}$_with respect to_the coordinate system $(x^i,y^i)$_are $$\begin{aligned} \left(u^i(x),\frac{\partial_u^i}{\partial_x^j}(x) y^j\right). \end{aligned}$$ Now the associated differential operator for $\mathbf{u}^{\#}\in\mathrm{Vect}(TM)$ that acts on smooth_functions_on $TM$_will_be_$$u^i(x)\frac{\partial}{\partial x^i} + \frac{\partial u^i}{\partial_x^j}(x)y^j \frac{\partial}{\partial y^i}.$$ The crucial_property of_complete lifts is that the assignment $\mathbf{u}\mapsto\mathbf{u}^{\#}$ is_a_*Lie algebra homomorphism*_$\mathrm{Vect}(M)\rightarrow\mathrm{Vect}(TM)$ with respect to the commutator bracket of vector_fields on $M$ and $TM$ respectively._Symbolically this is_$$\begin{aligned} \label{homo1} __ \left([\mathbf{u},\mathbf{v}]_M\right)^{\#} =_\left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM},\end{aligned}$$ which can be checked by_direct computation or otherwise. (The subscripts_indicate the space on which the vector_fields live.) A proof sketch can_be found in appendix **\[app:diff\]**. This_immediately implies_that vector fields $\mathbf{u}\in\mathrm{Vect}(M)$ on_$M$ can act_on functions_$f\in C^\infty(TM)$ on_$TM$ via $f\mapsto \mathbf{u}^{\#}\cdot f$, and_moreover this action_is a Lie algebra representation: $$\begin{aligned} \left[\mathbf{u}^{\#},\mathbf{v}^{\#}\right]_{TM}_\cdot_f = \mathbf{u}^{\#}\cdot\mathbf{v}^{\#}\cdot_f_-_\mathbf{v}^{\#}\cdot\mathbf{u}^{\#}\cdot f_= \left([\mathbf{u},\mathbf{v}]_M\right)^{\#} \cdot_f.\end{aligned}$$_Thus we_can_add this representation of $\mathrm{Vect}(M)$ on_$

\}. \end{aligned}$$ Then $\Sigma_s$ is the image of the projection $\pi_1: {\widetilde \Sigma_s} {\longrightarrow}G(t,S_d)$. Each fibre of the projection $\pi_2: {\widetilde \Sigma_s} {\longrightarrow}U_s $ is of dimension $t(s-t)$, hence $\dim {\widetilde \Sigma_s} =s+t(s-t)$. [[proof.]{}]{}Each $\Lambda \in \pi_2^{-1}({{\underline L}})$ is a subspace of the (finite) union $$\bigcup\limits_{\ell_{{{\underline \alpha}}} \leqslant s} W({{\underline \alpha}},{{\underline L}}),$$ hence it is a subspace of one of the $W$. Now use the fact that $\dim G(t,W)$ equals $t (\dim W -t)$. [[$\Box$]{}]{} Thus we have a na[" i]{}ve estimate $$\dim \, \Sigma_s \le \min\, \{\, \underbrace{s+t(s-t)}_{N_1}, \,\underbrace{t(d-t+1)}_{N_2} \, \}. \label{dimsigma.s}$$ In (\[dimsigma.s\]), we have an equality. \[theorem.dimsigma\] [[proof.]{}]{} Assume $N_1 < N_2$. We will exhibit a level Hilbert function ${{\underline h}}$ of type $(t,d)$ such that $h_s = s$ and $\dim {\mathcal L}^\circ({{\underline h}}) = N_1$. Then by construction, for each $\Lambda \in {\mathcal L}^\circ({{\underline h}})$ there is a nonzero form in $R_s$ which is apolar $\Lambda$. This form defines a secant $(s-1)$-plane containing ${{\mathbf P}}\Lambda$. Hence ${\mathcal L}^\circ({{\underline h}}) \subseteq \Sigma_s$, which forces $\dim \Sigma_s = N_1$. Let $m$ be the unique integer such that $(m+1)t > s \ge mt$. Then $N_1 < N_2$ forces $s \le d-m$. Define a sequence ${{\underline h}}$ by $$h_i = \begin{cases} i+1 & \text{for $0 \le i \le s-

\ }. \end{aligned}$$ Then $ \Sigma_s$ is the image of the projection $ \pi_1: { \widetilde \Sigma_s } { \longrightarrow}G(t, S_d)$. Each fibre of the projection $ \pi_2: { \widetilde \Sigma_s } { \longrightarrow}U_s $ is of property $ t(s - t)$, therefore $ \dim { \widetilde \Sigma_s } = s+t(s - t)$. [ [ proof.]{}]{}Each $ \Lambda \in \pi_2^{-1}({{\underline L}})$ is a subspace of the (finite) union $ $ \bigcup\limits_{\ell_{{{\underline \alpha } } } \leqslant s } W({{\underline \alpha}},{{\underline L}}),$$ hence it is a subspace of one of the $ W$. nowadays practice the fact that $ \dim G(t, W)$ equals $ t (\dim W -t)$. [ [ $ \Box$ ] { } ] { } Thus we have a na [ " i]{}ve calculate $ $ \dim \, \Sigma_s \le \min\, \{\, \underbrace{s+t(s - t)}_{N_1 }, \,\underbrace{t(d - t+1)}_{N_2 } \, \ }. \label{dimsigma.s}$$ In (\[dimsigma.s\ ]), we give birth an equality. \[theorem.dimsigma\ ] [ [ proof. ] { } ] { } Assume $ N_1 < N_2$. We will exhibit a level Hilbert routine $ { { \underline h}}$ of type $ (t, d)$ such that $ h_s = s$ and $ \dim { \mathcal L}^\circ({{\underline henry } }) = N_1$. Then by structure, for each $ \Lambda \in { \mathcal L}^\circ({{\underline h}})$ there is a nonzero form in $ R_s$ which is apolar $ \Lambda$. This human body defines a secant $ (s-1)$-plane containing $ { { \mathbf P}}\Lambda$. therefore $ { \mathcal L}^\circ({{\underline h } }) \subseteq \Sigma_s$, which forces $ \dim \Sigma_s = N_1$. Let $ m$ be the singular integer such that $ (m+1)t > s \ge mt$. Then $ N_1 < N_2 $ forces $ s \le d - m$. Define a succession $ { { \underline h}}$ by $ $ h_i = \begin{cases } i+1 & \text{for $ 0 \le i \le s-

\}. \end{aligned}$$ Then $\Sigma_s$ ls the image of jhw projxction $\li_1: {\widetklde \Sigma_s} {\longrightarrow}G(t,D_d)$. Wach dibre of the projectiov $\pi_2: {\widenilde \Signa_s} {\oingrightarckw}U_s $ is of djlensnoi $t(s-t)$, hence $\dim {\widetilde \Sigma_s} =s+t(s-t)$. [[prmow.]{}]{}Ecch $\Lambda \in \pi_2^{-1}({{\underline L}})$ is a subfpace og hhe (finite) unijn $$\bpgsup\ljmits_{\ell_{{{\underline \alpha}}} \leqslant s} W({{\undernine \alpha}},{{\undrrline L}}),$$ hence it is a subdpacf of one of the $W$. Jow use the face that $\dim G(t,D)$ equals $t (\dim W -t)$. [[$\Box$]{}]{} Jhus we have a na[" i]{}ve estimate $$\dkm \, \Sngma_s \le \nib\, \{\, \ktderbrace{s+t(w-t)}_{N_1}, \,\underbrace{t(d-t+1)}_{N_2} \, \}. \ldbel{dimxigma.s}$$ In (\[dimsinma.s\]), xe hqve an equality. \[theorxm.dimsigma\] [[proof.]{}]{} Assuie $N_1 < N_2$. Wa cill exhibit a level Yiobert funwtiov ${{\unaermiie g}}$ of tjpe $(t,d)$ such tgat $h_s = s$ abd $\dim {\mathcal L}^\cirv({{\ugeerline h}}) = N_1$. Fhen br sonstruction, for each $\Lambda \in {\mathcal L}^\cjrc({{\underline h}})$ there is a nonzero form in $R_s$ which is apolar $\Lambda$. This form defines a secant $(s-1)$-plane wontamnkng ${{\mwgybv P}}\Lambda$. Hence ${\mathcal L}^\circ({{\underline h}}) \subsqfea \Xigma_s$, which fjrces $\dim \Sogla_x = N_1$. Let $m$ be tfe unique integer such that $(m+1)t > s \de mt$. Then $N_1 < G_2$ fotces $s \le d-m$. Define a sequebce ${{\underlinv h}}$ vy $$h_i = \begin{cases} i+1 & \text{for $0 \ue i \le s-

\}. \end{aligned}$$ Then $\Sigma_s$ is the image projection {\widetilde \Sigma_s} Each fibre of {\longrightarrow}U_s is of dimension hence $\dim {\widetilde =s+t(s-t)$. [[proof.]{}]{}Each $\Lambda \in \pi_2^{-1}({{\underline L}})$ a subspace of the (finite) union $$\bigcup\limits_{\ell_{{{\underline \alpha}}} \leqslant s} W({{\underline \alpha}},{{\underline L}}),$$ it is a subspace of one of the $W$. Now use the fact $\dim equals (\dim -t)$. [[$\Box$]{}]{} Thus we have a na[" i]{}ve estimate $$\dim \, \Sigma_s \le \min\, \{\, \underbrace{s+t(s-t)}_{N_1}, \, \}. \label{dimsigma.s}$$ In (\[dimsigma.s\]), we have an \[theorem.dimsigma\] [[proof.]{}]{} Assume $N_1 N_2$. We will exhibit a Hilbert ${{\underline h}}$ type such $h_s = s$ $\dim {\mathcal L}^\circ({{\underline h}}) = N_1$. Then by construction, for each $\Lambda \in {\mathcal L}^\circ({{\underline h}})$ there a nonzero $R_s$ which apolar This defines a secant ${{\mathbf P}}\Lambda$. Hence ${\mathcal L}^\circ({{\underline h}}) forces $\dim \Sigma_s = N_1$. Let $m$ be unique integer that $(m+1)t > s \ge mt$. $N_1 < N_2$ forces $s \le d-m$. Define sequence ${{\underline h}}$ by $$h_i = \begin{cases} i+1 & \text{for $0 \le i \le s-

\}. \end{aligned}$$ Then $\Sigma_s$ is the Image of the ProjeCtiOn $\pI_1: {\wIdetIlde \sigma_s} {\longrighTArroW}G(t,S_d)$. Each fibre of the proJectiOn $\PI_2: {\widETiLde \SiGma_s} {\lonGRiGHTarRoW}U_S $ is Of DImEnsioN $t(s-T)$, hence $\dIm {\widetildE \SiGmA_s} =s+t(s-t)$. [[proof.]{}]{}EACh $\lambda \in \pi_2^{-1}({{\UndErline L}})$ is a suBspAce of tHe (FinITe) uniOn $$\bIgcup\Limits_{\ELl_{{{\undeRline \alphA}}} \lEQslant S} w({{\underlINE \aLpha}},{{\Underline L}}),$$ hence it IS a SUbspace of one of The $W$. NoW uSE tHE FacT thAt $\dim G(t,W)$ eqUaLs $t (\diM w -t)$. [[$\Box$]{}]{} ThUS wE HAVe a NA[" i]{}ve estimate $$\dIm \, \Sigma_s \le \mIN\, \{\, \unDerbraCe{S+t(s-T)}_{n_1}, \,\underBrace{T(d-T+1)}_{n_2} \, \}. \laBel{dimsigma.S}$$ In (\[dImsigma.s\]), wE have aN EqualitY. \[Theorem.DimsigMa\] [[pRooF.]{}]{} AssUMe $n_1 < N_2$. we wIlL ExhIBiT a lEVel hilbert fUnCtIon ${{\unDerlINE H}}$ Of tyPe $(t,D)$ sucH that $H_s = s$ and $\dim {\mathCal l}^\cirC({{\UndErlinE h}}) = N_1$. ThEn by CoNstruCtion, fOr eacH $\LAmbda \in {\mathcal L}^\Circ({{\Underline H}})$ thErE is A nOnzerO Form in $r_s$ wHicH is apolAr $\LambdA$. thiS fORM DeFines a secant $(s-1)$-plane CoNTAiNing ${{\mathBf P}}\LamBDa$. heNCe ${\mathcaL L}^\CirC({{\undERLine h}}) \SubsETeQ \Sigma_s$, wHich foRCeS $\dIm \Sigma_S = N_1$. let $m$ be ThE unIquE inteGEr suCh that $(M+1)t > s \ge mt$. THen $N_1 < N_2$ FOrces $s \le d-m$. DefiNE a sequence ${{\undERlINE h}}$ BY $$h_i = \bEgiN{cases} i+1 & \text{For $0 \lE I \le s-

\}. \end{aligned}$$ Th en $\Sigma _s$ i s t heim ageof t he projection$ \pi_ 1: {\widetilde \Sigma_ s} {\ lo n grig h ta rrow} G(t,S_d ) $. Eac hfi bre o f t he pr oje ction $ \pi_2: {\w ide ti lde \Sigma_s } { \longright arr ow}U_s $ isofdimens io n $ t (s-t) $,hence $\dim {\wide tilde \Si gm a _s} =s + t(s-t)$ . [ [pro of.]{}]{}Each $\L a mb d a \in \pi_2^{- 1}({{\ un d er l i neL}} )$ is a su bs paceo f the ( f in i t e ) u n ion $$\bigcup \limits_{\e l l_{ {{\und er lin e \alph a}}}\l e qsl ant s} W({{ \und erline \a lpha}} , {{\unde r line L} }),$$hen ceit i s a s ubs pa c e o f o neo f t he $W$.No wuse t he f a c t that $\ dimG(t,W )$ equals $t(\d im W -t) $. [[ $\Box $]{} ]{ } Th us wehaveana[" i]{}ve est imat e $$\dim\,\S igm a_ s \le \mi n\, \{ \, \und erbrace { s+t (s - t ) }_ {N_1}, \,\under br a c e{ t(d-t+1) }_{N_2 } \ ,\ }. \la be l{d imsi g m a.s}$ $ I n ( \[dimsig ma.s\] ) ,we have a nequali ty . \ [th eorem . dims igma\] [[proo f.]{} ] {} Assume $N_1 < N_2$. We wi l le x hi b it a le vel Hilbert fun c tion ${{ \ un der l ine h }}$ o ft yp e $(t,d)$ such that$h _s = s $ and $\dim {\math cal L}^\ci r c ( {{\under line h} } ) = N_1$. Then by c onstructio n , for ea ch $\ Lambda \ in {\math c a l L}^\ci rc( {{\ und erl i n eh}})$ there i s a no nz ero for m i n $R_s$ wh ich is ap ol ar $\Lamb da$. Thi sfo rm d efi nes a secant $ (s -1) $- pla ne co n tainin g ${{ \mat hb fP }}\ Lambda$ . H e n ce $ {\ ma thca l L }^ \circ ({{\ u nde rline h }}) \subs ete q \Si gm a_ s$, whi ch forces $\d im \Sigma_s=N_1 $. Le t $m$ be t he unique integer sucht hat $(m +1) t > s \ge mt$. The n $ N_1 <N_2 $ force s $s \ le d- m$ . D e f ine a s eq uen ce ${{\under l i neh}}$by $$h _i = \b egin{cases} i+1 &\ tex t{for $0 \lei \ le s -

\}._ _ \end{aligned}$$ Then $\Sigma_s$ is_the image_of_the projection_$\pi_1:_{\widetilde \Sigma_s} {\longrightarrow}G(t,S_d)$. Each_fibre of the_projection $\pi_2: {\widetilde \Sigma_s}_{\longrightarrow}U_s $ is_of_dimension $t(s-t)$, hence $\dim {\widetilde \Sigma_s} =s+t(s-t)$. [[proof.]{}]{}Each $\Lambda \in \pi_2^{-1}({{\underline L}})$ is a subspace_of_the (finite)_union_$$\bigcup\limits_{\ell_{{{\underline_\alpha}}} \leqslant s} W({{\underline \alpha}},{{\underline_L}}),$$ hence it is a_subspace of_one of the $W$. Now use the fact_that_$\dim G(t,W)$ equals_$t (\dim W -t)$. [[$\Box$]{}]{} Thus we have a na["_i]{}ve estimate $$\dim \, \Sigma_s \le_ \min\,_\{\,_\underbrace{s+t(s-t)}_{N_1},_ \,\underbrace{t(d-t+1)}_{N_2} \,_\}. \label{dimsigma.s}$$ In (\[dimsigma.s\]), we have_an equality. \[theorem.dimsigma\] [[proof.]{}]{} Assume $N_1 <_N_2$. We will exhibit a level Hilbert_function ${{\underline h}}$ of type $(t,d)$_such that $h_s = s$_and $\dim_{\mathcal L}^\circ({{\underline h}}) = N_1$._Then by construction,_for each_$\Lambda \in {\mathcal_L}^\circ({{\underline h}})$ there is a nonzero_form in $R_s$_which is apolar $\Lambda$. This form_defines_a secant $(s-1)$-plane_containing_${{\mathbf_P}}\Lambda$. Hence_${\mathcal L}^\circ({{\underline h}})_\subseteq_\Sigma_s$, which_forces_$\dim \Sigma_s = N_1$. Let $m$ be_the_unique integer such that $(m+1)t > s_\ge mt$. Then $N_1_<_N_2$ forces $s \le_d-m$. Define a sequence ${{\underline_h}}$ by $$h_i = \begin{cases} i+1 &_\text{for $0_\le i_\le s-

============================= Using the well known Gauss summation formula $$\label{Gauss} _2\phi_1\left[ \begin{array}{c} a,b\\ c \end{array};q, c/ab \right]=\frac{(c/a,c/b;q)_{\infty}}{(c,c/ab;q)_{\infty}},$$ where $|c/ab|<1$, we get a semi-finite form of Ramanujan’s summation of the general $_1\psi_1$, $$\label{Ran} _{1}\psi_1 \left[ \begin{array}{l} a\\ b \end{array};q,z\right]=\sum_{k=-\infty}^{\infty}\frac{(a;q)_k}{(b;q)_k}z^k= \frac{(q;q)_{\infty}(b/a;q)_{\infty}(az;q)_{\infty}(q/az;q)_{\infty}} {(b;q)_{\infty}(q/a;q)_{\infty}(z;q)_{\infty}(b/az;q)_{\infty}},$$ where $|b/a|<|z|<1$. For $|z|<1$, the following identity holds: \[theo\] $$\label{r-f} \sum_{k=-m}^{\infty}\frac{(a;q)_k(bq^{m}/az;q)_k}{(q^{1+m};q)_k(b;q)_k}z^k =\frac{(q;q)_m(q/az;q)_m}{(q/a;q)_m(b/az;q)_m} \frac{(b/a;q)_{\infty}(az;q)_{\infty}}{(b;q)_{\infty}(z;q)_{\infty}}.$$ [*Proof.*]{} The left hand side of (\[r-f\]) can be rewritten as $$\begin{aligned} \lefteqn{\sum_{k=0}^{\infty}\frac{(a;q)_{k-m}(bq^{m}/az;q)_{k-m}} {(q^{1+m};q)_{k-m}(b;q)_{k-m}}z^{k-m}}\\[6pt] &=&z^{-m}\frac{(a;q)_{-m}(bq^{m}/az;q)_{-m}}{(q

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = Using the well known Gauss summation formula $ $ \label{Gauss } _ 2\phi_1\left [ \begin{array}{c } a, b\\ vitamin c \end{array};q, c / ab \right]=\frac{(c / a, c / b;q)_{\infty}}{(c, c / ab;q)_{\infty}},$$ where $ |c / ab|<1 $, we make a semi - finite form of Ramanujan ’s summation of the general $ _ 1\psi_1 $, $ $ \label{Ran } _ { 1}\psi_1 \left [ \begin{array}{l } a\\ b \end{array};q, z\right]=\sum_{k=-\infty}^{\infty}\frac{(a;q)_k}{(b;q)_k}z^k= \frac{(q;q)_{\infty}(b / a;q)_{\infty}(az;q)_{\infty}(q / az;q)_{\infty } } { (b;q)_{\infty}(q / a;q)_{\infty}(z;q)_{\infty}(b / az;q)_{\infty}},$$ where $ |b / a|<|z|<1$. For $ |z|<1 $, the postdate identity holds: \[theo\ ] $ $ \label{r - f } \sum_{k=-m}^{\infty}\frac{(a;q)_k(bq^{m}/az;q)_k}{(q^{1+m};q)_k(b;q)_k}z^k = \frac{(q;q)_m(q / az;q)_m}{(q / a;q)_m(b / az;q)_m } \frac{(b / a;q)_{\infty}(az;q)_{\infty}}{(b;q)_{\infty}(z;q)_{\infty}}.$$ [ * Proof. * ] { } The leftover hand side of (\[r - f\ ]) can be rewritten as $ $ \begin{aligned } \lefteqn{\sum_{k=0}^{\infty}\frac{(a;q)_{k - m}(bq^{m}/az;q)_{k - m } } { (q^{1+m};q)_{k - m}(b;q)_{k - m}}z^{k - m}}\\[6pt ] & = & z^{-m}\frac{(a;q)_{-m}(bq^{m}/az;q)_{-m}}{(q

============================= Usijg the well known Gauss rummation formula $$\laben{Gauss} _2\lhi_1\left[ \bdgin{array}{c} a,b\\ c \end{array};q, c/ab \rijht]=\feac{(c/a,x/b;q)_{\infty}}{(c,c/ab;q)_{\infty}},$$ whefe $|c/ab|<1$, we get a swmi-fmnite form of Ramanujan’s summaflon oy vhe general $_1\psi_1$, $$\label{Ran} _{1}\pvi_1 \left[ \begin{arsah}{l} c\\ b \end{array};q,z\right]=\sum_{k=-\infty}^{\infty}\frac{(a;z)_k}{(b;q)_k}z^k= \grwc{(q;q)_{\infty}(b/a;q)_{\insty}(ae;q)_{\igfty}(s/az;q)_{\infty}} {(b;q)_{\infty}(q/a;q)_{\infty}(z;q)_{\infty}(b/az;s)_{\infty}},$$ xhere $|b/a|<|z|<1$. For $|z|<1$, yhe following identity holfs: \[tjeo\] $$\label{r-f} \sum_{k=-m}^{\invty}\frac{(a;q)_k(bw^{m}/az;z)_j}{(q^{1+m};q)_k(b;q)_k}z^k =\frxc{(q;q)_m(q/az;q)_m}{(q/a;q)_m(b/az;q)_m} \frad{(b/a;q)_{\infty}(az;q)_{\infty}}{(b;q)_{\infty}(z;q)_{\infty}}.$$ [*Oroof.*]{} The left yabd dhde of (\[r-f\]) cen be gewritten as $$\nvgin{alicned} \lefyeqn{\sum_{k=0}^{\infty}\fvac{(a;q)_{n-m}(bw^{m}/az;q)_{k-m}} {(q^{1+m};q)_{k-m}(b;q)_{k-m}}z^{k-m}}\\[6pt] &=&v^{-m}\frac{(a;q)_{-m}(bq^{m}/az;q)_{-m}}{(q

============================= Using the well known Gauss summation _2\phi_1\left[ a,b\\ c c/ab \right]=\frac{(c/a,c/b;q)_{\infty}}{(c,c/ab;q)_{\infty}},$$ where form Ramanujan’s summation of general $_1\psi_1$, $$\label{Ran} \left[ \begin{array}{l} a\\ b \end{array};q,z\right]=\sum_{k=-\infty}^{\infty}\frac{(a;q)_k}{(b;q)_k}z^k= \frac{(q;q)_{\infty}(b/a;q)_{\infty}(az;q)_{\infty}(q/az;q)_{\infty}} where $|b/a|<|z|<1$. For $|z|<1$, the following identity holds: \[theo\] $$\label{r-f} \sum_{k=-m}^{\infty}\frac{(a;q)_k(bq^{m}/az;q)_k}{(q^{1+m};q)_k(b;q)_k}z^k =\frac{(q;q)_m(q/az;q)_m}{(q/a;q)_m(b/az;q)_m} \frac{(b/a;q)_{\infty}(az;q)_{\infty}}{(b;q)_{\infty}(z;q)_{\infty}}.$$ The left hand side of (\[r-f\]) can be rewritten as $$\begin{aligned} \lefteqn{\sum_{k=0}^{\infty}\frac{(a;q)_{k-m}(bq^{m}/az;q)_{k-m}} {(q^{1+m};q)_{k-m}(b;q)_{k-m}}z^{k-m}}\\[6pt]

============================= Using the well known Gauss sumMation formUla $$\laBel{gauSs} _2\Phi_1\lEft[ \bEgin{array}{c} a,b\\ c \eND{arrAy};q, c/ab \right]=\frac{(c/a,c/b;q)_{\inFty}}{(c,c/Ab;Q)_{\InftY}},$$ WhEre $|c/aB|<1$, we get a SEmI-FIniTe FoRm oF RAMaNujan’S suMmation Of the generAl $_1\pSi_1$, $$\Label{Ran} _{1}\psi_1 \lEFt[ \Begin{array}{L} a\\ b \End{array};q,z\riGht]=\Sum_{k=-\inFtY}^{\inFTy}\fraC{(a;q)_K}{(b;q)_k}z^K= \frac{(q;Q)_{\Infty}(b/A;q)_{\infty}(az;Q)_{\iNFty}(q/az;Q)_{\Infty}} {(b;q)_{\INFtY}(q/a;q)_{\Infty}(z;q)_{\infty}(b/az;q)_{\iNFtY}},$$ Where $|b/a|<|z|<1$. For $|z|<1$, thE folloWiNG iDENtiTy hOlds: \[theo\] $$\laBeL{r-f} \suM_{K=-m}^{\infty}\FRaC{(A;Q)_K(bq^{M}/Az;q)_k}{(q^{1+m};q)_k(b;q)_k}z^k =\Frac{(q;q)_m(q/az;q)_M}{(Q/a;q)_M(b/az;q)_m} \FrAc{(b/A;Q)_{\infty}(Az;q)_{\inFtY}}{(B;q)_{\iNfty}(z;q)_{\infty}}.$$ [*prooF.*]{} The left hAnd sidE Of (\[r-f\]) can BE rewritTen as $$\bEgiN{alIgneD} \LeFtEqn{\SuM_{K=0}^{\inFTy}\FraC{(A;q)_{k-M}(bq^{m}/az;q)_{k-M}} {(q^{1+M};q)_{K-m}(b;q)_{k-M}}z^{k-m}}\\[6PT] &=&Z^{-M}\Frac{(A;q)_{-m}(Bq^{m}/aZ;q)_{-m}}{(q

========================== === Using thewel l k no wn G auss summation for m ula$$\label{Gauss} _2\phi _1\le ft [ \be g in {arra y}{c} a , b\ \ c \ en d{ arr ay } ;q , c/a b \ right]= \frac{(c/a ,c/ b; q)_{\infty}} { (c ,c/ab;q)_{ \in fty}},$$ whe re$|c/ab |< 1$, we ge t a semi -finit e formof Ramanu ja n ’s sum m ation o f th e ge neral $_1\psi_1$, $$ \ label{Ran} _{1 }\psi_ 1\ le f t [ \ beg in{array}{ l} a\\b \end{a r ra y } ; q,z \ right]=\sum_{ k=-\infty}^ { \in fty}\f ra c{( a ;q)_k} {(b;q )_ k }z^ k= \frac{(q ;q)_ {\infty}( b/a;q) _ {\infty } (az;q)_ {\inft y}( q/a z;q) _ {\ in fty }} {(b ; q) _{\ i nft y}(q/a;q )_ {\ infty }(z; q ) _ { \inf ty} (b/a z;q)_ {\infty}},$$whe re $ | b/a |<|z| <1$. For $ |z|<1 $, the foll ow ing identity ho lds: \[theo\] $$ \l abe l{ r-f}\ sum_{k =-m }^{ \infty} \frac{( a ;q) _k ( b q ^{ m}/az;q)_k}{(q^{1+ m} ; q )_ k(b;q)_k }z^k = \ fr ac { (q;q)_m( q/ az; q)_m } { (q/a; q)_m ( b/ az;q)_m} \frac { (b /a ;q)_{\i nf ty}(az ;q )_{ \in fty}} { (b;q )_{\in fty}(z;q )_{\i n fty}}.$$ [*Pr o of.*]{} The l e ft h an d sid e o f (\[r-f\]) can be r ewri t te n a s $$\b egin{ al i gn e d} \lefteqn{\sum_{k =0 }^{\in fty}\ frac{(a;q)_{k -m}(bq^{m} / a z ;q)_{k-m }} { ( q^ { 1+m};q)_{k-m}( b;q)_ {k-m}}z^{k - m}}\\[6p t] &= &z^{-m}\ frac{(a;q ) _ {-m}(bq^ {m} /az ;q) _{- m } }{ (q

============================= Using the_well known_Gauss summation formula $$\label{Gauss} _2\phi_1\left[ \begin{array}{c} a,b\\ c \end{array};q,_c/ab \right]=\frac{(c/a,c/b;q)_{\infty}}{(c,c/ab;q)_{\infty}},$$ where_$|c/ab|<1$,_we get_a_semi-finite form of_Ramanujan’s summation of_the general $_1\psi_1$, $$\label{Ran} _{1}\psi_1_\left[ \begin{array}{l} a\\ b \end{array};q,z\right]=\sum_{k=-\infty}^{\infty}\frac{(a;q)_k}{(b;q)_k}z^k= \frac{(q;q)_{\infty}(b/a;q)_{\infty}(az;q)_{\infty}(q/az;q)_{\infty}} {(b;q)_{\infty}(q/a;q)_{\infty}(z;q)_{\infty}(b/az;q)_{\infty}},$$ where $|b/a|<|z|<1$. For_$|z|<1$,_the following identity holds: \[theo\] $$\label{r-f} \sum_{k=-m}^{\infty}\frac{(a;q)_k(bq^{m}/az;q)_k}{(q^{1+m};q)_k(b;q)_k}z^k =\frac{(q;q)_m(q/az;q)_m}{(q/a;q)_m(b/az;q)_m} \frac{(b/a;q)_{\infty}(az;q)_{\infty}}{(b;q)_{\infty}(z;q)_{\infty}}.$$ [*Proof.*]{} The left hand side of (\[r-f\]) can be_rewritten_as $$\begin{aligned} \lefteqn{\sum_{k=0}^{\infty}\frac{(a;q)_{k-m}(bq^{m}/az;q)_{k-m}} {(q^{1+m};q)_{k-m}(b;q)_{k-m}}z^{k-m}}\\[6pt] &=&z^{-m}\frac{(a;q)_{-m}(bq^{m}/az;q)_{-m}}{(q

or $\{n-2,n\}$, and hence for each of these edges, $f_0(\operatorname{\mathrm{lk}}(e,\Delta(I))= (2n-5)-1+2<2n-3$. On the other hand, the links of $\{2,3\}$ in $B(I)$ and $-B(I)$ are both $1$-simplices, and so $f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=(2n-5)-2+4=2n-3$. Therefore, for edges $e\subseteq V_n$, the link $\operatorname{\mathrm{lk}}(e, \Delta(I))$ has $2n-3$ vertices if and only if $e=\pm \{2,3\}$. Now, if $e=\{i, n+1\}$, then $f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=f_0(\operatorname{\mathrm{lk}}(i, \partial B(I)))= f_0(\operatorname{\mathrm{lk}}(i, B(I)))$. By definition of $B(I)$, $f_0(\operatorname{\mathrm{lk}}(2, B(I)))=n-1$. Indeed, the vertex set of $\operatorname{\mathrm{lk}}(2,B(I))$ consists of all vertices (but $2$ itself) that appear in the nodes of the middle column of Figure \[figure: tree\], i.e., it is $\{1\}\cup [3,n]$. This shows that $f_0(\operatorname{\mathrm{lk}}(\{2, n+1\}, \Delta(I)))=n-1$. To see that other edges containing $n+1$ have “short" links observe that the vertex set of $\operatorname{\mathrm{lk}}(1,B(I))$ does not contain $-n+3, -n+4$. For any $i\in[3,n]$, the vertex set of $\operatorname{\mathrm{lk}}(i, B(I))$ contains at most six positive vertices, namely, $\{1,2, i-2,i-1,i+1,i+2\}\cap [n]$. Finally, for any negative vertex $i\in V_n$, the vertex set of $\operatorname{\mathrm{lk}}(i,B(I))$ misses $3$ and $4$. Hence we conclude that

or $ \{n-2,n\}$, and hence for each of these edges, $ f_0(\operatorname{\mathrm{lk}}(e,\Delta(I))= (2n-5)-1 + 2<2n-3$. On the other hand, the connection of $ \{2,3\}$ in $ B(I)$ and $ -B(I)$ are both $ 1$-simplices, and therefore $ f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=(2n-5)-2 + 4=2n-3$. Therefore, for edges $ e\subseteq V_n$, the link $ \operatorname{\mathrm{lk}}(e, \Delta(I))$ have $ 2n-3 $ vertices if and only if $ e=\pm \{2,3\}$. immediately, if $ e=\{i, n+1\}$, then $ f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=f_0(\operatorname{\mathrm{lk}}(i, \partial B(I)))= f_0(\operatorname{\mathrm{lk}}(i, B(I)))$. By definition of $ B(I)$, $ f_0(\operatorname{\mathrm{lk}}(2, B(I)))=n-1$. Indeed, the vertex set up of $ \operatorname{\mathrm{lk}}(2,B(I))$ consists of all vertex (but $ 2 $ itself) that appear in the nodes of the center column of Figure \[figure: tree\ ], i.e., it is $ \{1\}\cup [ 3,n]$. This shows that $ f_0(\operatorname{\mathrm{lk}}(\{2, n+1\ }, \Delta(I)))=n-1$. To see that early edges containing $ n+1 $ have “ short " links observe that the vertex set of $ \operatorname{\mathrm{lk}}(1,B(I))$ does not contain $ -n+3, -n+4$. For any $ i\in[3,n]$, the vertex sic of $ \operatorname{\mathrm{lk}}(i, B(I))$ contains at most six positive vertex, namely, $ \{1,2, i-2,i-1,i+1,i+2\}\cap [ n]$. Finally, for any negative vertex $ i\in V_n$, the vertex set of $ \operatorname{\mathrm{lk}}(i, B(I))$ misses $ 3 $ and $ 4$. therefore we conclude that

or $\{n-2,n\}$, and hence for each on these edges, $f_0(\okeeatorneme{\mathdm{lk}}(e,\Delga(I))= (2n-5)-1+2<2n-3$. On the other hand, the lunks if $\{2,3\}$ in $B(I)$ and $-B(I)$ are buth $1$-simplpces, and wo $f_0(\iperatornamx{\jathrm{lk}}(e, \Delfw(I)))=(2n-5)-2+4=2n-3$. Chxrefore, for edggs $e\subseteq V_n$, the link $\o[efacorname{\mathrm{lk}}(e, \Delta(I))$ has $2n-3$ verticqs if amd only if $e=\pm \{2,3\}$. Njw, ig $e=\{i, h+1\}$, then $f_0(\operatorname{\mathrm{lk}}(e, \Deltz(I)))=f_0(\operetorname{\mathrm{ll}}(i, \partial B(I)))= f_0(\operatornamf{\matjrm{lk}}(i, B(I)))$. By definltion of $B(I)$, $f_0(\opqeatorname{\matfrm{lk}}(2, B(I)))=n-1$. Pudeed, the vgrtex set of $\operatorname{\mathrm{ly}}(2,B(I))$ cpnsists of aol ggrtices (but $2$ itsejf) that appear in tha nodes of the middle conumb of Figure \[figure: trxe\], i.e., it is $\{1\}\cup [3,n]$. Thys shows dhct $f_0(\operatorname{\mathrn{lj}}(\{2, n+1\}, \Dglta(I)))=t-1$. To wee thzt ofher efgea containihg $n+1$ have “syort" links observe uhae the vertex sst of $\jpqratorname{\mathrm{lk}}(1,B(I))$ does not contain $-n+3, -n+4$. Ror any $i\in[3,n]$, the vertex set of $\operatorname{\mwthrm{lk}}(i, F(I))$ contains at most six positive vertices, namely, $\{1,2, i-2,i-1,i+1,i+2\}\rao [n]$. Flnalut, vor any negative vertex $i\in V_n$, the vertex set kf $\oieratorname{\mathrm{jk}}(i,B(I))$ missex $3$ amq $4$. Hence we cuncludz tgat

or $\{n-2,n\}$, and hence for each of $f_0(\operatorname{\mathrm{lk}}(e,\Delta(I))= On the hand, the links $-B(I)$ both $1$-simplices, and $f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=(2n-5)-2+4=2n-3$. Therefore, edges $e\subseteq V_n$, the link $\operatorname{\mathrm{lk}}(e, has $2n-3$ vertices if and only if $e=\pm \{2,3\}$. Now, if $e=\{i, n+1\}$, $f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=f_0(\operatorname{\mathrm{lk}}(i, \partial B(I)))= f_0(\operatorname{\mathrm{lk}}(i, B(I)))$. By definition of $B(I)$, $f_0(\operatorname{\mathrm{lk}}(2, B(I)))=n-1$. Indeed, vertex of consists all vertices (but $2$ itself) that appear in the nodes of the middle column of Figure tree\], i.e., it is $\{1\}\cup [3,n]$. This shows $f_0(\operatorname{\mathrm{lk}}(\{2, n+1\}, \Delta(I)))=n-1$. To that other edges containing $n+1$ “short" observe that vertex of does not contain -n+4$. For any $i\in[3,n]$, the vertex set of $\operatorname{\mathrm{lk}}(i, B(I))$ contains at most six positive vertices, namely, i-2,i-1,i+1,i+2\}\cap [n]$. any negative $i\in the set of $\operatorname{\mathrm{lk}}(i,B(I))$ and $4$. Hence we conclude that

or $\{n-2,n\}$, and hence for each of thesE edges, $f_0(\opeRatorNamE{\maThRm{lk}}(E,\DelTa(I))= (2n-5)-1+2<2n-3$. On the otheR Hand, The links of $\{2,3\}$ in $B(I)$ and $-B(I)$ are Both $1$-sImPLiceS, AnD so $f_0(\oPeratorNAmE{\MAthRm{Lk}}(E, \DeLtA(i)))=(2n-5)-2+4=2N-3$. TherEfoRe, for edGes $e\subsetEq V_N$, tHe link $\operatORnAme{\mathrm{lK}}(e, \DElta(I))$ has $2n-3$ verTicEs if anD oNly IF $e=\pm \{2,3\}$. NOw, iF $e=\{i, n+1\}$, tHen $f_0(\opERatornAme{\mathrm{Lk}}(E, \delta(I)))=F_0(\OperatoRNAmE{\matHrm{lk}}(i, \partial B(I)))= f_0(\oPErATorname{\mathrm{lK}}(i, B(I)))$. By DeFInITIon Of $B(i)$, $f_0(\operatorNaMe{\matHRm{lk}}(2, B(I)))=n-1$. iNdEED, The VErtex set of $\opeRatorname{\maTHrm{Lk}}(2,B(I))$ coNsIstS Of all vErticEs (BUt $2$ iTself) that apPear In the nodeS of the MIddle coLUmn of FiGure \[fiGurE: trEe\], i.e., IT iS $\{1\}\cUp [3,n]$. thIS shOWs ThaT $F_0(\opEratornaMe{\MaThrm{lK}}(\{2, n+1\}, \DeLTA(i)))=N-1$. To sEe tHat oTher eDges containinG $n+1$ hAve “sHOrt" Links ObserVe thAt The veRtex seT of $\opErAtorname{\mathrm{lK}}(1,B(I))$ dOes not conTaiN $-n+3, -N+4$. FoR aNy $i\in[3,N]$, The verTex Set Of $\operaTorname{\MAthRm{LK}}(I, b(I))$ Contains at most six pOsITIvE verticeS, namelY, $\{1,2, I-2,i-1,I+1,i+2\}\CAp [n]$. FinalLy, For Any nEGAtive VertEX $i\In V_n$, the vErtex sET oF $\oPeratorNaMe{\mathRm{Lk}}(i,b(I))$ mIsses $3$ ANd $4$. HeNce we cOnclude tHat

or $\{n-2,n\}$, and hence for eachof th ese ed ge s, $ f_0( \operatorname{ \ math rm{lk}}(e,\Delta(I))=(2n-5 )- 1 +2<2 n -3 $. On the ot h er h and ,th e l in k sof $\ {2, 3\}$ in $B(I)$ an d $ -B (I)$ are bot h $ 1$-simplic es, and so $f_0 (\o perato rn ame { \math rm{ lk}}( e, \De l ta(I)) )=(2n-5)- 2+ 4 =2n-3$ . There f o re , fo r edges $e\subset e qV _n$, the link$\oper at o rn a m e{\ mat hrm{lk}}(e ,\Delt a (I))$ h a s$ 2 n -3$ vertices if a nd only if$ e=\ pm \{2 ,3 \}$ . Now, if $ e= \ {i, n+1\}$, th en $ f_0(\oper atorna m e{\math r m{lk}}( e, \De lta (I) ))=f _ 0( \o per at o rna m e{ \ma t hrm {lk}}(i, \ pa rtial B(I ) ) ) = f_0 (\o pera torna me{\mathrm{lk }}( i, B ( I)) )$. B y def init io n of$B(I)$ , $f_ 0( \operatorname{\ math rm{lk}}(2 , B (I ))) =n -1$.I ndeed, th e v ertex s et of $ \ ope ra t o r na me{\mathrm{lk}}(2, B( I ) )$ consist s of a l lve r tices (b ut $2 $ it s e lf) t hata pp ear in t he nod e sof the mi dd le col um n o f F igure \[fi gure:tree\],i.e., it is $\{1\}\c u p [3,n]$. Thi s s h o ws that $f _0(\operato rnam e {\ma thrm { lk }}( \ {2, n +1\}, \ D el t a(I)))=n-1$. To see t hat ot her e dges containi ng $n+1$ h a v e “short" lin k so bserve that th e ver tex set of $\operat ornam e{\mathr m{lk}}(1, B ( I))$ doe s n otcon tai n $- n+3, -n+4$. F o r any $ i\in[3, n]$ , the v ert exset of $ \operator name{\ma th rm {l k} }(i , B(I ) )$ conta in s a tmos t six positi ve ve rtic es ,n ame ly, $\{ 1 ,2 , i-2, i- 1, i+1, i+2 \} \cap[n]$ . Fi nally,for any n ega t iveve rt ex $i\i n V_n$, the v er tex set of $ \op erator n a me{\math rm{lk}}(i,B(I))$ misses $3$ and $4 $. He ncewe conclu dethat

or_$\{n-2,n\}$, and_hence for each of_these edges,_$f_0(\operatorname{\mathrm{lk}}(e,\Delta(I))=_(2n-5)-1+2<2n-3$. On_the_other hand, the_links of $\{2,3\}$_in $B(I)$ and $-B(I)$_are both $1$-simplices,_and_so $f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=(2n-5)-2+4=2n-3$. Therefore, for edges $e\subseteq V_n$, the link $\operatorname{\mathrm{lk}}(e, \Delta(I))$ has $2n-3$ vertices_if_and only_if_$e=\pm_\{2,3\}$. Now, if $e=\{i, n+1\}$, then_$f_0(\operatorname{\mathrm{lk}}(e, \Delta(I)))=f_0(\operatorname{\mathrm{lk}}(i, \partial B(I)))= f_0(\operatorname{\mathrm{lk}}(i,_B(I)))$. By_definition of $B(I)$, $f_0(\operatorname{\mathrm{lk}}(2, B(I)))=n-1$. Indeed, the vertex_set_of $\operatorname{\mathrm{lk}}(2,B(I))$ consists_of all vertices (but $2$ itself) that appear in_the nodes of the middle column_of Figure \[figure:_tree\],_i.e.,_it is $\{1\}\cup [3,n]$._This shows that $f_0(\operatorname{\mathrm{lk}}(\{2, n+1\}, \Delta(I)))=n-1$._To see that other edges containing_$n+1$ have “short" links observe that the_vertex set of $\operatorname{\mathrm{lk}}(1,B(I))$ does not_contain $-n+3, -n+4$. For any_$i\in[3,n]$, the_vertex set of $\operatorname{\mathrm{lk}}(i, B(I))$_contains at most_six positive_vertices, namely, $\{1,2,_i-2,i-1,i+1,i+2\}\cap [n]$. Finally, for any negative_vertex $i\in V_n$,_the vertex set of $\operatorname{\mathrm{lk}}(i,B(I))$ misses_$3$_and $4$. Hence_we_conclude_that

and integrate over $t$ to find $E_{\rm fb}$ as $$E_{\rm fb} = \dfrac{3}{5} \left( \dfrac{4 \pi \rho_{\rm c} }{3} \right)^{1/3} G (M_{\rm c})^{5/3},$$ where we have assumed that all of the gravitational potential energy is liberated instantaneously. We can then study this feedback energy as a fraction of the total energy of the protoplanet from Equation \[eq:binding\], $$\dfrac{E_{\rm fb}}{E_{\rm tot}} = 1.3 \left(\dfrac{\rho_{\rm c}}{5 \rm gcm^{-3}}\right)^{1/3} \left(\dfrac{M_{\rm c}}{20 M_{\oplus}}\right)^{5/3} \left( \dfrac{r_{\rm p}}{1 \rm AU}\right) \left( \dfrac{5 M_{J}}{M_{p}}\right)^2 \label{eq:feedback}$$ or alternatively, $$\dfrac{E_{\rm fb}}{E_{\rm tot}} = \left( \dfrac{M_{\rm c}}{M_{\rm p}} \right)^2 \dfrac{ r_{\rm p}}{r_{\rm c}}. \label{eq:fb_simple}$$ Let us consider a protoplanet with $M_{\rm p} = 5M_J$, $ r_{\rm p}$ = 1 AU, and a rocky core with mass $M_{\rm c}=M_\oplus$ and $\rho_{\rm c}$ = 5 gcm$^{-3}$. Equation \[eq:feedback\] gives us a value of $\sim$ 0.008, far too small to disrupt the protoplanet. However, if we allow the core to grow to 20 $M_\oplus$ [Jupiter’s core mass is now believed to be $\sim 10 M_{\oplus}$; @WahlEtal17; @Nettelmann17; @GuillotEtal18], we get a value of 1.3. This result is exciting since it shows us that reasonably modest cores have the potential to destroy young GI protoplanets. Figure \[fig:an\_feedback\] illustrates Equation \[eq:feedback\] for a 5 $M_J$ protoplanet with two different assumptions for the initial radius. In the remainder of this paper we demonstrate the

and integrate over $ t$ to find $ E_{\rm fb}$ as $ $ E_{\rm fb } = \dfrac{3}{5 } \left (\dfrac{4 \pi \rho_{\rm c } } { 3 } \right)^{1/3 } G (M_{\rm c})^{5/3},$$ where we have assumed that all of the gravitational likely department of energy is release instantaneously. We can then study this feedback department of energy as a fraction of the total department of energy of the protoplanet from Equation \[eq: binding\ ], $ $ \dfrac{E_{\rm fb}}{E_{\rm tot } } = 1.3 \left(\dfrac{\rho_{\rm c}}{5 \rm gcm^{-3}}\right)^{1/3 } \left(\dfrac{M_{\rm c}}{20 M_{\oplus}}\right)^{5/3 } \left (\dfrac{r_{\rm p}}{1 \rm AU}\right) \left (\dfrac{5 M_{J}}{M_{p}}\right)^2 \label{eq: feedback}$$ or alternatively, $ $ \dfrac{E_{\rm fb}}{E_{\rm tot } } = \left (\dfrac{M_{\rm c}}{M_{\rm p } } \right)^2 \dfrac { r_{\rm p}}{r_{\rm c } }. \label{eq: fb_simple}$$ Let us regard a protoplanet with $ M_{\rm phosphorus } = 5M_J$, $ r_{\rm p}$ = 1 AU, and a rocky core with mass $ M_{\rm c}=M_\oplus$ and $ \rho_{\rm c}$ = 5 gcm$^{-3}$. Equation \[eq: feedback\ ] gives us a value of $ \sim$ 0.008, far besides small to disrupt the protoplanet. However, if we leave the core to grow to 20 $ M_\oplus$ [ Jupiter ’s core mass is now believed to be $ \sim 10 M_{\oplus}$; @WahlEtal17; @Nettelmann17; @GuillotEtal18 ], we get a value of 1.3. This consequence is exciting since it shows us that reasonably modest kernel have the potential to destroy young gilbert protoplanets. Figure \[fig: an\_feedback\ ] illustrate Equation \[eq: feedback\ ] for a 5 $ M_J$ protoplanet with two different assumptions for the initial radius. In the end of this paper we demonstrate the

anf integrate over $t$ to fikd $E_{\rm fb}$ as $$E_{\rm yv} = \dfcac{3}{5} \lert( \dfrac{4 \pi \rho_{\rm c} }{3} \right)^{1/3} G (M_{\rm c})^{5/3},$$ whxre qe hace assumed that all of the gravptational pottntial energy is liberated instzktanemnsly. We can thek study thiv feedback enesgh cs a fraction of the total energy of the prptlplanet from Ezuatpog \[eq:gpnbing\], $$\dfrac{E_{\rm fb}}{E_{\rm tot}} = 1.3 \left(\dfdac{\rho_{\rk c}}{5 \rm gcm^{-3}}\rignt)^{1/3} \left(\dfrac{M_{\rm c}}{20 M_{\oplus}}\rihht)^{5/3} \peft( \dfrac{r_{\rm p}}{1 \rm AU}\right) \ledt( \dseac{5 M_{J}}{M_{p}}\righg)^2 \label{eq:fetdyack}$$ or altetnatively, $$\dfrac{E_{\rm fb}}{E_{\rm tot}} = \lefg( \dfrcc{M_{\rm c}}{M_{\rm k}} \dihvt)^2 \dfrac{ r_{\rn p}}{r_{\ri c}}. \label{ca:fb_sim[le}$$ Let is consider a iroto'lanwt with $M_{\rm p} = 5M_J$, $ r_{\rm p}$ = 1 AU, and a rocki core witv jass $M_{\rm c}=M_\oplus$ qne $\rho_{\tm c}$ = 5 gco$^{-3}$. Eqjatjoi \[es:feedbwck\] gives us z value of $\wim$ 0.008, far too small uo qpxrupt the prktoplagee. However, if we allow the core to grow uo 20 $M_\kplus$ [Jupiter’s core masw is now believed to he $\sim 10 M_{\jplus}$; @WahlEtal17; @Nettelmann17; @GuillotEtal18], we get a vanue oh 1.3. Thnw resjot is exciting since it shows us that reasonablr mpdvst cores have thc potential to desyrly ioung GI protoklanets. Fifure \[fig:an\_feedback\] illusttates Wquation \[tq:feecback\] for a 5 $M_J$ protoplanet with two diyfeeent assumptions flr the initnal racius. On the remainder of thir paler we demojstrate tgd

and integrate over $t$ to find $E_{\rm $$E_{\rm = \dfrac{3}{5} \dfrac{4 \pi \rho_{\rm c})^{5/3},$$ we have assumed all of the potential energy is liberated instantaneously. We then study this feedback energy as a fraction of the total energy of protoplanet from Equation \[eq:binding\], $$\dfrac{E_{\rm fb}}{E_{\rm tot}} = 1.3 \left(\dfrac{\rho_{\rm c}}{5 \rm gcm^{-3}}\right)^{1/3} c}}{20 \left( p}}{1 AU}\right) \left( \dfrac{5 M_{J}}{M_{p}}\right)^2 \label{eq:feedback}$$ or alternatively, $$\dfrac{E_{\rm fb}}{E_{\rm tot}} = \left( \dfrac{M_{\rm c}}{M_{\rm p}} \right)^2 r_{\rm p}}{r_{\rm c}}. \label{eq:fb_simple}$$ Let us consider a with $M_{\rm p} = $ r_{\rm p}$ = 1 and rocky core mass c}=M_\oplus$ $\rho_{\rm c}$ = gcm$^{-3}$. Equation \[eq:feedback\] gives us a value of $\sim$ 0.008, far too small to disrupt the protoplanet. if we core to to $M_\oplus$ core mass is to be $\sim 10 M_{\oplus}$; @WahlEtal17; get a value of 1.3. This result is since it us that reasonably modest cores have potential to destroy young GI protoplanets. Figure \[fig:an\_feedback\] Equation \[eq:feedback\] for a 5 $M_J$ protoplanet with two different assumptions for the initial radius. remainder of this paper demonstrate the

and integrate over $t$ to find $E_{\rM fb}$ as $$E_{\rm fb} = \Dfrac{3}{5} \LefT( \dfRaC{4 \pi \rHo_{\rm C} }{3} \right)^{1/3} G (M_{\rm c})^{5/3},$$ wheRE we hAve assumed that all of the GraviTaTIonaL PoTentiAl energY Is LIBerAtEd InsTaNTaNeousLy. WE can theN study this FeeDbAck energy as a FRaCtion of the TotAl energy of thE prOtoplaNeT frOM EquaTioN \[eq:biNding\], $$\dFRac{E_{\rm Fb}}{E_{\rm tot}} = 1.3 \lEfT(\Dfrac{\rHO_{\rm c}}{5 \rm gCM^{-3}}\RiGht)^{1/3} \lEft(\dfrac{M_{\rm c}}{20 M_{\opluS}}\RiGHt)^{5/3} \left( \dfrac{r_{\rm P}}{1 \rm AU}\rIgHT) \lEFT( \dfRac{5 m_{J}}{M_{p}}\right)^2 \lAbEl{eq:fEEdback}$$ oR AlTERNatIVely, $$\dfrac{E_{\rm fB}}{E_{\rm tot}} = \left( \DFraC{M_{\rm c}}{M_{\Rm P}} \riGHt)^2 \dfraC{ r_{\rm p}}{R_{\rM C}}. \laBel{eq:fb_simpLe}$$ LeT us considEr a proTOplanet WIth $M_{\rm p} = 5m_J$, $ r_{\rm p}$ = 1 aU, aNd a RockY CoRe WitH mASs $M_{\RM c}=m_\opLUs$ aNd $\rho_{\rm c}$ = 5 GcM$^{-3}$. EQuatiOn \[eq:FEEDBack\] GivEs us A valuE of $\sim$ 0.008, far too sMalL to dISruPt the ProtoPlanEt. howevEr, if we Allow ThE core to grow to 20 $M_\oPlus$ [jupiter’s cOre MaSs iS nOw belIEved to Be $\sIm 10 M_{\Oplus}$; @WaHlEtal17; @NETteLmANN17; @guIllotEtal18], we get a valUe OF 1.3. thIs result Is exciTInG sINce it shoWs Us tHat rEASonabLy moDEsT cores haVe the pOTeNtIal to deStRoy youNg gI pRotOplanETs. FiGure \[fiG:an\_feedbAck\] ilLUstrates EquatiON \[eq:feedback\] foR A 5 $M_j$ PRoTOplaNet With two diffErenT AssuMptiONs For THe iniTial rAdIUs. iN the remainder of this PaPer we dEmonsTrate the

and integrate over $t$ to find $E_{ \rm f b}$ as $$E_ {\rm fb} = \dfrac { 3}{5 } \left( \dfrac{4 \pi \rho _{ \ rm c } } {3} \ right)^ { 1/ 3 } G(M _{ \rm c } )^ {5/3} ,$$ where we have a ssu me d that all o f t he gravita tio nal potentia l e nergyis li b erate d i nstan taneou s ly. We can then s t udy th i s feedb a c kener gy as a fractiono ft he total energ y of t he pr o t opl ane t from Equ at ion \ [ eq:bind i ng \ ] , $ $ \dfrac{E_{\rm fb}}{E_{\r m to t}} = 1 .3\ left(\ dfrac {\ r ho_ {\rm c}}{5\rmgcm^{-3}} \right ) ^{1/3}\ left(\d frac{M _{\ rmc}}{ 2 0M_ {\o pl u s}} \ ri ght ) ^{5 /3} \lef t( \ dfrac {r_{ \ r m p}}{ 1 \ rm A U}\ri ght) \left( \ dfr ac{5 M_{ J}}{M _{p}} \rig ht )^2 \ label{ eq:fe ed back}$$ or alt erna tively, $$\ df rac {E _{\rm fb}}{E _{\ rmtot}} = \left( \df ra c { M _{ \rm c}}{M_{\rm p}} \ r i gh t)^2 \df rac{ r _ {\ rm p}}{r_{\ rm c} }. \lab el{e q :f b_simple }$$ L e tus consid er a pro to pla net with $M_{ \rm p} = 5M_J$ , $ r _ {\rm p}$ = 1 A U , and a rocky co r e w i th m ass $M_{\rm c} =M_\ o plus $ an d $ \rh o _{\rm c}$=5 g c m$^{-3}$. Equation\[ eq:fee dback \] gives us a value of$ \ s im$ 0.00 8, f a rt oo small to di srupt the proto p lanet. H oweve r, if we allow th e core togro w t o 2 0 $ M _ \o plus$ [Jupite r ’ s co re mass i s n ow beli eve d t o b e $ \s im 10 M_{ \oplus}$ ;@W ah lE tal 17; @ N ettelman n1 7;@G uil lotEt a l18],we ge t ava lu e of 1.3. T h is r esul tis exc iti ng sinc e it sho ws us t hat reaso nab l y mo de st coreshave the pote nt ial to des tr oyyoungG I protopl anets. Figure \[fig:an\ _ feedbac k\] illu stra tes Equat ion \[eq: fee d back\] for a 5 $M _J $ p r o topla n e twit htwo differ e n t a ssump ti onsfor the initial radius. I n th e remainder o f t hisp a pe r w e d e mon st r ate t he

and_integrate over_$t$ to find $E_{\rm_fb}$ as $$E_{\rm_fb}_= _\dfrac{3}{5}_ \left( \dfrac{4_\pi \rho_{\rm c}_}{3} \right)^{1/3} G (M_{\rm_c})^{5/3},$$ where we have_assumed_that all of the gravitational potential energy is liberated instantaneously. We can then study_this_feedback energy_as_a_fraction of the total energy_of the protoplanet from Equation_\[eq:binding\], $$\dfrac{E_{\rm fb}}{E_{\rm_tot}} = 1.3 \left(\dfrac{\rho_{\rm c}}{5 \rm gcm^{-3}}\right)^{1/3} \left(\dfrac{M_{\rm_c}}{20_M_{\oplus}}\right)^{5/3} \left( \dfrac{r_{\rm_p}}{1 \rm AU}\right) \left( \dfrac{5 M_{J}}{M_{p}}\right)^2 \label{eq:feedback}$$ or alternatively, $$\dfrac{E_{\rm fb}}{E_{\rm tot}}_= \left( \dfrac{M_{\rm c}}{M_{\rm p}} \right)^2_\dfrac{ r_{\rm p}}{r_{\rm_c}}. __ \label{eq:fb_simple}$$ Let us_consider a protoplanet with $M_{\rm p}_= 5M_J$, $ r_{\rm p}$ = 1_AU, and a rocky core with mass_$M_{\rm c}=M_\oplus$ and $\rho_{\rm c}$ =_5 gcm$^{-3}$. Equation \[eq:feedback\] gives_us a_value of $\sim$ 0.008, far_too small to_disrupt the_protoplanet. However, if_we allow the core to grow_to 20 $M_\oplus$_[Jupiter’s core mass is now believed_to_be $\sim 10_M_{\oplus}$;_@WahlEtal17;_@Nettelmann17; @GuillotEtal18],_we get a_value_of 1.3._This_result is exciting since it shows_us_that reasonably modest cores have the potential_to destroy young GI_protoplanets._Figure \[fig:an\_feedback\] illustrates Equation_\[eq:feedback\] for a 5 $M_J$_protoplanet with two different assumptions for_the initial_radius. In_the remainder of this paper we demonstrate the

the vector meson $B_c^*$ allows us to ignore here the distinction between the scalar and vector mesons. The CDF collaboration recently announced the discovery of the $B_{c} $ system [@abe198; @abe298]. Their announcement was based on careful analysis of 20 events with characteristic three-charged particle tracks expected from the decay $B_{c}^{\pm } \rightarrow J/\psi + l^{\pm} + X. $ The CDF collaboration determined values for both the mass and the lifetime of the ground state, $$\nonumber M(B_{c}) = 6400 \pm 390 \pm 130 \; {\rm MeV}, \hspace{0.5in} \tau(B_{c}) = 0.46^{+0.18}_{-0.16} \pm 0.03 \; {\rm ps}.$$ Allowing that either the c quark or the $\bar{\rm b}$ quark can decay while the other is simply a spectator, or that the c quark and the $\bar{\rm b }$ can annihilate into an intermediate vector boson, the weak decay width of the ground state of the $B_{c} $ is the sum of three terms, namely, $$\Gamma (B_{c} \rightarrow X) = \Gamma (\bar{b} \rightarrow X) + \Gamma ( c\rightarrow X ) + \Gamma(\mbox{annih}).$$ Thus our potential model calculation of section \[bcbound\] yields $\tau(B_{c}) = 0.36 \pm 0.05 \; ps$, which agrees with the CDF result. Our calculation of the mass of $B_{c} $ described below in section \[bcbound\] is also in good agreement with the CDF measurement. This lifetime allows $B_{c} $ mesons with sufficient transverse velocity to escape the beam centroid and thus produce a secondary tri-lepton vertex, which serves as a distinctive feature for identifying the decay. It is noteworthy that the lifetimes of all $B_{c} $ states below threshold are much longer than their counterparts in charmonium or the upsilon system because flavor conservation prevents decays through gluon annihilation diagrams. In a QGP environment at RHIC the formation of this ‘stable’ exotic $B_{c}$ meson is facilitated by the ready availability of charmed quarks. We explore here the production enhancement of $B_c$ with a view at both a study of its properties and as diagnostic tool

the vector meson $ B_c^*$ allows us to ignore here the distinction between the scalar and vector mesons. The CDF collaboration recently announce the discovery of the $ B_{c } $ arrangement   [ @abe198; @abe298 ]. Their announcement was based on careful analysis of 20 events with characteristic three - charge particle tracks expect from the decay $ B_{c}^{\pm } \rightarrow J/\psi + l^{\pm } + X. $ The CDF collaboration specify values for both the batch and the lifetime of the ground department of state, $ $ \nonumber M(B_{c }) = 6400 \pm 390 \pm 130 \; { \rm MeV }, \hspace{0.5 in } \tau(B_{c }) = 0.46^{+0.18}_{-0.16 } \pm 0.03 \; { \rm ps}.$$ Allowing that either the c quark cheese or the $ \bar{\rm b}$ quark can decay while the early is simply a spectator pump, or that the deoxycytidine monophosphate quark and the $ \bar{\rm b } $ can annihilate into an intermediate vector boson, the unaccented decay width of the ground state of the $ B_{c } $ is the kernel of three terms, namely, $ $ \Gamma (B_{c } \rightarrow X) = \Gamma (\bar{b } \rightarrow X) + \Gamma ( c\rightarrow X) + \Gamma(\mbox{annih}).$$ therefore our potential mannequin calculation of section \[bcbound\ ] yields $ \tau(B_{c }) = 0.36 \pm 0.05 \; ps$, which agrees with the CDF resultant role. Our calculation of the mass of $ B_{c } $ described below in section \[bcbound\ ] is also in good agreement with the CDF measurement. This lifetime allows $ B_{c } $ mesons with sufficient transverse velocity to escape the beam centroid and thus produce a secondary tri - lepton vertex, which serves as a classifiable feature of speech for identifying the decay. It is noteworthy that the lifetimes of all $ B_{c } $ states below brink are much longer than their counterparts in charmonium or the upsilon arrangement because flavor conservation prevents decays through gluon annihilation diagrams. In a QGP environment at RHIC the constitution of this ‘ stable ’ exotic $ B_{c}$ meson is facilitated by the ready availability of charmed quark cheese. We explore here the production enhancement of $ B_c$ with a view at both a study of its property and as diagnostic instrument

thf vector meson $B_c^*$ allows us to ignore hgrw the vistincfion betdeen the scalar and vector mxsonw. The CDF collaboration recdntly annlunced tye dmscovery of the $U_{d} $ systci [@abe198; @wbe298]. Chxir announcemenj was based mn careful ananyrid of 20 events with characteristic thwee-charbef particle trasks txpqctes from the decay $B_{c}^{\pm } \rightarrow N/\psi + l^{\km} + X. $ The CDF colkaboration determined valufs flr both the mass ajd the lifejjme if the grouna state, $$\nonumber M(B_{c}) = 6400 \km 390 \pm 130 \; {\rm MeV}, \hspace{0.5in} \tau(B_{c}) = 0.46^{+0.18}_{-0.16} \pm 0.03 \; {\rm ps}.$$ Alloqibg hvat either vhe c zuark or the $\bar{\rm b}$ quark van decay whilc the otyer is simply a spectetor, or that the c qoark and tve $\bar{\rm b }$ can annuhulate intm an untdrmtdiete vectog bkson, the wsak decay wudth of the ground xtwnr of the $B_{c} $ is thq fum of three terms, namely, $$\Gamma (B_{c} \righuarros X) = \Gamma (\bar{b} \rightareow X) + \Gamma ( c\rigjtarrow X ) + \Gamma(\mbox{annih}).$$ Thus our potential model calculdtion uf wegtiov \[bfbound\] yields $\tau(B_{c}) = 0.36 \pm 0.05 \; ps$, which agrees wieg uhe CDF result. Ouv calculation of tne msfs of $B_{c} $ deszribed bemow in section \[bcblund\] is also in good wgrerment with the CDF measuremwnt. This lifenime allows $B_{c} $ mesons with suffieient jransvrrse velocity to escape the beam centrlid and tgjs produce a secundsrf tri-lepuun vertex, which sqrves as e discinctive feajure fow identifylng tma decay. It is noteaorthi that the lifetlmes of all $B_{c} $ states below thcxshold are muvh lotger thau theiv counterparts yn charmonium pr the bpsilov system bvcause flevor conservwtion preventv decays throngh gluon annuhilqtion dkxgrams. In a QGP environmvnu at RHIC rhe formation of tmis ‘sjagle’ exotic $B_{c}$ mzrin is facilitatrd cy ehv rxady weailability mf cfarord quxrks. We expoure nere the production anhahcement of $B_c$ with a view at both a ftudy of its lroperties and as fiagnmstmc took

the vector meson $B_c^*$ allows us to the between the and vector mesons. the of the $B_{c} system [@abe198; @abe298]. announcement was based on careful analysis 20 events with characteristic three-charged particle tracks expected from the decay $B_{c}^{\pm } J/\psi + l^{\pm} + X. $ The CDF collaboration determined values for both mass the of ground state, $$\nonumber M(B_{c}) = 6400 \pm 390 \pm 130 \; {\rm MeV}, \hspace{0.5in} \tau(B_{c}) = \pm 0.03 \; {\rm ps}.$$ Allowing that either c quark or the b}$ quark can decay while other simply a or the quark and the b }$ can annihilate into an intermediate vector boson, the weak decay width of the ground state the $B_{c} the sum three namely, (B_{c} \rightarrow X) (\bar{b} \rightarrow X) + \Gamma ( + \Gamma(\mbox{annih}).$$ Thus our potential model calculation of \[bcbound\] yields = 0.36 \pm 0.05 \; ps$, agrees with the CDF result. Our calculation of mass of $B_{c} $ described below in section \[bcbound\] is also in good agreement with measurement. This lifetime allows $ mesons with transverse to the centroid and produce a secondary tri-lepton vertex, which serves as a distinctive feature identifying the decay. It is noteworthy that the lifetimes of $ below threshold are longer than their counterparts charmonium the upsilon system because prevents through In QGP at RHIC the formation this ‘stable’ exotic $B_{c}$ meson facilitated by the ready explore here the production enhancement of $B_c$ with view at both a study of its and as diagnostic tool

the vector meson $B_c^*$ allows us tO ignore herE the dIstIncTiOn beTweeN the scalar and vECtor Mesons. The CDF collaboratIon reCeNTly aNNoUnced The discOVeRY Of tHe $b_{c} $ SysTeM [@AbE198; @abe298]. THeiR announCement was bAseD oN careful analYSiS of 20 events wIth CharacteristIc tHree-chArGed PArticLe tRacks ExpectED from tHe decay $B_{c}^{\Pm } \RIghtarROw J/\psi + l^{\PM} + x. $ THe CDf collaboration detERmINed values for boTh the mAsS AnD THe lIfeTime of the gRoUnd stATe, $$\nonumBEr m(b_{C}) = 6400 \Pm 390 \pM 130 \; {\Rm MeV}, \hspace{0.5in} \Tau(B_{c}) = 0.46^{+0.18}_{-0.16} \pm 0.03 \; {\rm ps}.$$ aLloWing thAt EitHEr the c Quark Or THe $\bAr{\rm b}$ quark cAn deCay while tHe otheR Is simplY A spectaTor, or tHat The C quaRK aNd The $\BaR{\Rm b }$ CAn AnnIHilAte into aN iNtErmedIate VECTOr boSon, The wEak deCay width of the GroUnd sTAte Of the $b_{c} $ is tHe suM oF threE terms, NamelY, $$\GAmma (B_{c} \rightarroW X) = \GaMma (\bar{b} \riGhtArRow x) + \GAmma ( c\RIghtarRow x ) + \GaMma(\mbox{Annih}).$$ ThUS ouR pOTENtIal model calculatioN oF SEcTion \[bcboUnd\] yieLDs $\TaU(b_{c}) = 0.36 \pm 0.05 \; ps$, whIcH agRees WITh the cDF rESuLt. Our calCulatiON oF tHe mass oF $B_{C} $ descrIbEd bEloW in seCTion \[BcbounD\] is also iN good AGreement with thE cDF measuremenT. thIS LiFEtimE alLows $B_{c} $ mesonS witH SuffIcieNT tRanSVerse VelocItY To EScape the beam centroiD aNd thus ProduCe a secondary tRi-lepton veRTEX, which seRves AS a DIstinctive featUre foR identifyiNG the decaY. It is NotewortHy that the LIFetimes oF alL $B_{c} $ StaTes BELoW threshold are MUCh loNgEr than tHeiR counteRpaRts In cHarMoNium or the Upsilon sYsTeM bEcAusE flavOR conservAtIon PrEveNts deCAys thrOugh gLuon AnNiHIlaTion diaGRaMS. in a QgP EnViroNmeNt At RHIc the FOrmAtion of This ‘stablE’ exOTic $B_{C}$ mEsOn is facIlitated by the ReAdy availabIlIty Of charMED quarks. WE explore here the productiON enhancEmeNt of $B_C$ witH a view at bOth A study Of iTS propeRties aNd as dIaGnoSTIc tooL

the vector meson $B_c^*$allows usto ig nor e h er e th e di stinction betw e en t he scalar and vector m esons .T he C D Fcolla boratio n r e c ent ly a nno un c ed thedis coveryof the $B_ {c} $ system [@ab e 19 8; @abe298 ].Their announ cem ent wa sbas e d oncar efulanalys i s of 2 0 eventswi t h char a cterist i c t hree -charged particle tr a cks expected f rom th ed ec a y $B _{c }^{\pm } \ ri ghtar r ow J/\p s i+ l^{ \ pm} + X. $ Th e CDF colla b ora tion d et erm i ned va luesfo r bo th the mass and the life time o f the gr o und sta te, $$ \no num berM (B _{ c}) = 640 0 \ pm3 90\pm 130\; { \rm M eV}, \ h s pace {0. 5in} \ta u(B_{c}) = 0. 46^ {+0. 1 8}_ {-0.1 6} \p m 0. 03 \; { \rm ps }.$$ A llowing that ei ther the c qu ark o r t he $\ba r {\rm b }$qua rk candecay w h ile t h e ot her is simply a sp ec t a to r, or th at the cqu a rk and t he $\ bar{ \ r m b } $ ca n a nnihilat e into an i ntermed ia te vec to r b oso n, th e wea k deca y widthof th e ground stateo f the $B_{c}$ i s th e sum of three term s, n a mely , $$ \ Ga mma (B_{c } \ri gh t ar r ow X) = \Gamma (\ba r{ b} \ri ghtar row X) + \Gam ma ( c \ r i ghtarrow X ) +\ Gamma(\mbox{an nih}) .$$ Thus o u r potent ial m odel cal culationo f section \[ bcb oun d\] y ie lds $\tau(B_{ c } ) =0. 36 \pm0.0 5 \; ps $,whi chagr ee s with th e CDF re su lt .Ou r c alcul a tion ofth e m as s o f $B_ { c} $ d escri bedbe lo w in sectio n \ [ b cbou nd \] isals oin go od a g ree ment wi th the CD F m e asur em en t. Thi s lifetime al lo ws $B_{c}$mes ons wi t h suffici ent transverse velocity to esca pethe b eamcentroidand thuspro d uce asecond ary t ri -le p t on ve r t ex , w hi ch servesa s adisti nc tive featur e for identifyingt hedecay. It isnot ewor t h ytha t t h e l if e tim e s of all $B_{c}$ states b el o wthresholda remu ch long er than thei r counte rparts in charmoni um ort h e u psilon sys tem beca use flavo r cons e rv ation pr events d eca ys th roughg luo n ann ihilat io n diag rams. In a QGP environment at RHIC th e form ation of this ‘st abl e ’ e xotic $B_ {c}$ meson isfac ili tated by the r eady av ail a bilit y of charmed q u ar ks. W eexplore her e t heprodu cti o n enha ncem ent of $B_c$ with a view at both a s t u dyofi ts p ro perties and as di ag n o stic too l

the_vector meson_$B_c^*$ allows us to_ignore here_the_distinction between_the_scalar and vector_mesons. The CDF_collaboration recently announced the_discovery of the_$B_{c}_$ system [@abe198; @abe298]. Their announcement was based on careful analysis of 20 events with_characteristic_three-charged particle_tracks_expected_from the decay $B_{c}^{\pm }_\rightarrow J/\psi + l^{\pm} +_X. $_The CDF collaboration determined values for both the_mass_and the lifetime_of the ground state, $$\nonumber M(B_{c}) = 6400 \pm 390_\pm 130 \; {\rm MeV}, \hspace{0.5in}_ \tau(B_{c}) = 0.46^{+0.18}_{-0.16}_\pm_0.03_\; {\rm ps}.$$ Allowing that_either the c quark or the_$\bar{\rm b}$ quark can decay while_the other is simply a spectator, or_that the c quark and the_$\bar{\rm b }$ can annihilate_into an_intermediate vector boson, the weak_decay width of_the ground_state of the_$B_{c} $ is the sum of_three terms, namely,_$$\Gamma (B_{c} \rightarrow X) = \Gamma_(\bar{b}_\rightarrow X) +_\Gamma_( _ _ c\rightarrow X_)_+ \Gamma(\mbox{annih}).$$_Thus_our potential model calculation of section_\[bcbound\]_yields $\tau(B_{c}) = 0.36 \pm 0.05 \;_ps$, which agrees with_the_CDF result. Our calculation_of the mass of $B_{c}_$ described below in section \[bcbound\]_is also_in good_agreement with the CDF measurement. This lifetime allows $B_{c} $ mesons with_sufficient transverse velocity to escape the_beam centroid and thus_produce a_secondary_tri-lepton vertex, which_serves_as a_distinctive feature for identifying the decay. It_is noteworthy_that the lifetimes of all $B_{c}_$ states below threshold_are_much longer than their counterparts in_charmonium or the upsilon system because_flavor conservation prevents decays through_gluon_annihilation_diagrams. In a QGP environment at_RHIC the formation of this ‘stable’_exotic $B_{c}$ meson_is facilitated by the ready availability of_charmed_quarks. We explore here the production_enhancement_of $B_c$ with a view at_both_a_study of its properties and_as diagnostic tool

bm}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}m_1-n_{j2}n_{11}+n_{j2}m_1}x_{-\bn_1+\bn_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}n_{j1}-m_2n_{11}+m_2n_{j1}}x_{-\bn_1+n_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ & +x_\bm^{-1}e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ =& x_\bm^{-3}a_{-3}+x_\bm^{-2}a_{-2}+x_\bm^{-1}a_{-1},\end{aligned}$$ where $$\begin{aligned} a_{-1}=&2i_1q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}+ e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ =& i_1(3+\mu-i_1-2i_2\cdots-2i_k)q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & + \sum_{j=2}^k\sum_{l=2}^k i_l (i_j-\delta_{j,l})q^{-n_{12}n_{j1}-n_{l2}n_{11}+n_{l2}n_{j1}}x_{-\bn_1+\bn_l+\bn_j}x_{\bn_j

bm}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & + x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}m_1 - n_{j2}n_{11}+n_{j2}m_1}x_{-\bn_1+\bn_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ & + x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}n_{j1}-m_2n_{11}+m_2n_{j1}}x_{-\bn_1+n_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ & + x_\bm^{-1}e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ = & x_\bm^{-3}a_{-3}+x_\bm^{-2}a_{-2}+x_\bm^{-1}a_{-1},\end{aligned}$$ where $ $ \begin{aligned } a_{-1}=&2i_1q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}+ e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ = & i_1(3+\mu - i_1 - 2i_2\cdots-2i_k)q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & + \sum_{j=2}^k\sum_{l=2}^k i_l (i_j-\delta_{j, l})q^{-n_{12}n_{j1}-n_{l2}n_{11}+n_{l2}n_{j1}}x_{-\bn_1+\bn_l+\bn_j}x_{\bn_j

bm}x_{\hn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^kl_jq^{-n_{12}m_1-n_{j2}n_{11}+n_{j2}m_1}x_{-\bn_1+\bn_l+\vm}x_{\bn_1}^{i_1}\rdots x_{\gn_j}^{i_j-1}\cdogs x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}n_{j1}-m_2n_{11}+m_2n_{o1}}x_{-\bn_1+b_j+\bm}x_{\vn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\cn_k}^{i_k}\\ & +x_\bm^{-1}v_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdits e_{\bn_k}^{i_k})\\ =& x_\bm^{-3}a_{-3}+x_\bm^{-2}a_{-2}+e_\gm^{-1}a_{-1},\end{aligned}$$ aherz $$\uegin{aligned} a_{-1}=&2i_1q^{-k_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdmts x_{\bn_k}^{i_k}+ e_{12}(-\bn_1)(f_{\bv_1}^{i_1}\edots x_{\bn_k}^{i_k})\\ =& i_1(3+\mu-i_1-2i_2\cdots-2i_k)q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cqots x_{\bm_k}^{l_k}\\ & + \sum_{j=2}^k\sum_{l=2}^k i_l (p_j-\qeltz_{j,l})q^{-n_{12}n_{j1}-n_{l2}n_{11}+n_{l2}n_{j1}}x_{-\bn_1+\bn_l+\bn_j}x_{\bn_j

bm}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}m_1-n_{j2}n_{11}+n_{j2}m_1}x_{-\bn_1+\bn_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ & x_{\bn_k}^{i_k}\\ +x_\bm^{-1}e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ x_\bm^{-3}a_{-3}+x_\bm^{-2}a_{-2}+x_\bm^{-1}a_{-1},\end{aligned}$$ where $$\begin{aligned} i_1(3+\mu-i_1-2i_2\cdots-2i_k)q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots & + \sum_{j=2}^k\sum_{l=2}^k (i_j-\delta_{j,l})q^{-n_{12}n_{j1}-n_{l2}n_{11}+n_{l2}n_{j1}}x_{-\bn_1+\bn_l+\bn_j}x_{\bn_j

bm}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^kI_jq^{-n_{12}m_1-n_{j2}n_{11}+n_{j2}M_1}x_{-\bn_1+\bN_j+\bM}x_{\bN_1}^{i_1}\CdotS x_{\bn_J}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ & +X_\Bm^{-2}\suM_{j=1}^ki_jq^{-n_{12}n_{j1}-m_2n_{11}+m_2n_{j1}}x_{-\bn_1+n_j+\bm}x_{\Bn_1}^{i_1}\cdOtS X_{\bn_j}^{I_J-1}\cDots x_{\Bn_k}^{i_k}\\ & +x_\bM^{-1}E_{12}(-\bN_1)(X_{\Bn_1}^{i_1}\CdOtS x_{\bN_k}^{I_K})\\ =& x_\Bm^{-3}a_{-3}+x_\bM^{-2}a_{-2}+x_\Bm^{-1}a_{-1},\end{aLigned}$$ wherE $$\beGiN{aligned} a_{-1}=&2i_1q^{-n_{11}N_{12}}X_{\bN_1}^{-1}x_{\bn_1}^{i_1}\cdots X_{\bn_K}^{i_k}+ e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdOts X_{\bn_k}^{i_k})\\ =& I_1(3+\mU-i_1-2i_2\CDots-2i_K)q^{-n_{11}N_{12}}x_{\bn_1}^{-1}x_{\Bn_1}^{i_1}\cdoTS x_{\bn_k}^{i_K}\\ & + \sum_{j=2}^k\sum_{L=2}^k I_L (i_j-\delTA_{j,l})q^{-n_{12}n_{j1}-N_{L2}N_{11}+n_{L2}n_{j1}}x_{-\Bn_1+\bn_l+\bn_j}x_{\bn_j

bm}x_{\bn_1}^{i_1}\cdots x _{\bn_k}^{ i_k}\ \ & +x _\ bm^{ -2}\ sum_{j=1}^ki_j q ^{-n _{12}m_1-n_{j2}n_{11}+ n_{j2 }m _ 1}x_ { -\ bn_1+ \bn_j+\ b m} x _ {\b n_ 1} ^{i _1 } \c dotsx_{ \bn_j}^ {i_j-1}\cd ots x _{\bn_k}^{i_ k }\ \ & +x_\bm ^{- 2}\sum_{j=1} ^ki _jq^{- n_ {12 } n_{j1 }-m _2n_{ 11}+m_ 2 n_{j1} }x_{-\bn_ 1+ n _j+\bm } x_{\bn_ 1 } ^{ i_1} \cdots x_{\bn_j}^ { i_ j -1}\cdots x_{\ bn_k}^ {i _ k} \ \ &+x_ \bm^{-1}e_ {1 2}(-\ b n_1)(x_ { \b n _ 1 }^{ i _1}\cdots x_{ \bn_k}^{i_k } )\\ =& x _\ bm^ { -3}a_{ -3}+x _\ b m^{ -2}a_{-2}+x _\bm ^{-1}a_{- 1},\en d {aligne d }$$ whe re $$\ beg in{ alig n ed }a_{ -1 } =&2 i _1 q^{ - n_{ 11}n_{12 }} x_ {\bn_ 1}^{ - 1 } x _{\b n_1 }^{i _1}\c dots x_{\bn_k }^{ i_k} + e_ {12}( -\bn_ 1)(x _{ \bn_1 }^{i_1 }\cdo ts x_{\bn_k}^{i_k })\\ =& i_1(3 +\m u- i_1 -2 i_2\c d ots-2i _k) q^{ -n_{11} n_{12}} x _{\ bn _ 1 } ^{ -1}x_{\bn_1}^{i_1} \c d o ts x_{\bn_ k}^{i_ k }\ \& + \sum_ {j =2} ^k\s u m _{l=2 }^ki _l (i_j-\d elta_{ j ,l }) q^{-n_{ 12 }n_{j1 }- n_{ l2} n_{11 } +n_{ l2}n_{ j1}}x_{- \bn_1 + \bn_l+\bn_j}x_ { \bn_j

bm}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ &_+x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}m_1-n_{j2}n_{11}+n_{j2}m_1}x_{-\bn_1+\bn_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots_x_{\bn_k}^{i_k}\\ & +x_\bm^{-2}\sum_{j=1}^ki_jq^{-n_{12}n_{j1}-m_2n_{11}+m_2n_{j1}}x_{-\bn_1+n_j+\bm}x_{\bn_1}^{i_1}\cdots x_{\bn_j}^{i_j-1}\cdots x_{\bn_k}^{i_k}\\ &_+x_\bm^{-1}e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ _=&_x_\bm^{-3}a_{-3}+x_\bm^{-2}a_{-2}+x_\bm^{-1}a_{-1},\end{aligned}$$ where_$$\begin{aligned} a_{-1}=&2i_1q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots_x_{\bn_k}^{i_k}+ e_{12}(-\bn_1)(x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k})\\ =&_i_1(3+\mu-i_1-2i_2\cdots-2i_k)q^{-n_{11}n_{12}}x_{\bn_1}^{-1}x_{\bn_1}^{i_1}\cdots x_{\bn_k}^{i_k}\\ & +_\sum_{j=2}^k\sum_{l=2}^k i_l (i_j-\delta_{j,l})q^{-n_{12}n_{j1}-n_{l2}n_{11}+n_{l2}n_{j1}}x_{-\bn_1+\bn_l+\bn_j}x_{\bn_j

}$]{} (90,29)[$\mbox{\boldmath$v$}$]{} (87,15)[${\bf D}$]{} (35,15)[$\mbox{\boldmath$u$}^{\ast}$]{} Consider how to determine the probability density $P_{(Z_1,Z_2)}$ for the representative point to reach $(Z_1,Z_2)$. Let $p(Z_1,Z_2,t,\Delta t)$ be the probability density for the point to be at $(Z_1,Z_2)$ at time $t$ after it is emiitted from S, and to be absorbed on B after time $t+\Delta t$. The probability density $P_{(Z_1,Z_2)}$ is obtained by integrating over the times: $$\label{eq: 3.1a} P_{(Z_1,Z_2)}=\int_0^\infty {\rm d}t\int_0^\infty {\rm d}\Delta t\ p(Z_1,Z_2,t,\Delta t)$$ Because the diffusion process is Markovian, we can write $p$ as a product of two functions: $p(Z_1,Z_2,t,\Delta t)=G(Z_1,Z_2,t)p_2(Z_1,Z_2,\Delta t)$, where $G(Z_1,Z_2,t)$ is the Green’s function (probability density to reach $(Z_1,Z_2)$ after time $t$), and $p_2(Z_1,Z_2,\Delta t)$ is the probability density for a particle released at $(Z_1,Z_2)$ to be absorbed on B after time $\Delta t$. From this definition of $p_2$, we have $$\label{eq: 3.1b} \int_0^\infty {\rm d}\Delta t\ p_2(Z_1,Z_2,\Delta t)=1$$ so that $$\label{eq: 3.2} P_{(Z_1,Z_2)}=\int_0^\infty {\rm d}t\ G(Z_1,Z_2,t) \.$$ Strictly speaking, $G(Z_1,Z_2,t)$ should be the Green’s function for reaching $(Z_1,Z_2)$ after time $t$ by a path

} $ ] { } (90,29)[$\mbox{\boldmath$v$}$ ] { } (87,15)[${\bf D}$ ] { } (35,15)[$\mbox{\boldmath$u$}^{\ast}$ ] { } Consider how to determine the probability density $ P_{(Z_1,Z_2)}$ for the representative point to strive $ (Z_1,Z_2)$. lease $ p(Z_1,Z_2,t,\Delta t)$ be the probability density for the point to be at $ (Z_1,Z_2)$ at time $ t$ after it is emiitted from S, and to be absorb on B after time $ t+\Delta t$. The probability density $ P_{(Z_1,Z_2)}$ is obtain by integrating over the times: $ $ \label{eq: 3.1a } P_{(Z_1,Z_2)}=\int_0^\infty { \rm d}t\int_0^\infty { \rm d}\Delta t\ p(Z_1,Z_2,t,\Delta t)$$ Because the dissemination process is Markovian, we can write $ p$ as a intersection of two functions: $ p(Z_1,Z_2,t,\Delta t)=G(Z_1,Z_2,t)p_2(Z_1,Z_2,\Delta t)$, where $ G(Z_1,Z_2,t)$ is the Green ’s function (probability concentration to reach $ (Z_1,Z_2)$ after time $ t$), and $ p_2(Z_1,Z_2,\Delta t)$ is the probability density for a particle released at $ (Z_1,Z_2)$ to be absorbed on B after time $ \Delta t$. From this definition of $ p_2 $, we suffer $ $ \label{eq: 3.1b } \int_0^\infty { \rm d}\Delta t\ p_2(Z_1,Z_2,\Delta t)=1$$ so that $ $ \label{eq: 3.2 } P_{(Z_1,Z_2)}=\int_0^\infty { \rm d}t\ G(Z_1,Z_2,t) \.$$ Strictly speaking, $ G(Z_1,Z_2,t)$ should be the Green ’s function for reach $ (Z_1,Z_2)$ after time $ t$ by a path

}$]{} (90,29)[$\mblx{\boldmath$v$}$]{} (87,15)[${\bf D}$]{} (35,15)[$\mbox{\bolamath$u$}^{\ast}$]{} Considgr how tm detedmine thd probability density $P_{(Z_1,Z_2)}$ foc thw repeesentative point to rdach $(Z_1,Z_2)$. Lvt $p(Z_1,Z_2,t,\Deota u)$ be the probabilmfy denslcy fod the 'oint to be at $(E_1,Z_2)$ at time $t$ after it is ekiktced from S, and to be absorbed on B aster tike $t+\Delta t$. The krobanylitg density $P_{(Z_1,Z_2)}$ is obtained by integdating mver the timex: $$\label{eq: 3.1a} P_{(Z_1,Z_2)}=\int_0^\infty {\rm f}t\inh_0^\infty {\rm d}\Delta t\ p(Z_1,Z_2,t,\Delta j)$$ Besquse the difwusion probzss is Markkvian, we can write $p$ as a produzt of two functuobs: $o(E_1,Z_2,t,\Delta t)=G(Z_1,V_2,t)p_2(Z_1,Z_2,\Dvlta t)$, where $N(D_1,Z_2,t)$ is dhe Grern’s function (pvobabmlitt density to reach $(Z_1,Z_2)$ after time $t$), and $k_2(Z_1,Z_2,\Delta t)$ ia the probability dwnsiti for a pxetizle rxlezsed ah $(Z_1,V_2)$ to be abskrbed on B qfter time $\Delta t$. Grjn this definifion os $[_2$, we have $$\label{eq: 3.1b} \int_0^\infty {\rm d}\Delta t\ p_2(Z_1,A_2,\Delta t)=1$$ so that $$\label{ew: 3.2} P_{(Z_1,Z_2)}=\int_0^\infty {\rm d}t\ G(E_1,Z_2,t) \.$$ Strictjy speaking, $G(Z_1,Z_2,t)$ should be the Green’s function fos rearhkng $(Z_1,D_2)$ afgwr time $t$ by a path

}$]{} (90,29)[$\mbox{\boldmath$v$}$]{} (87,15)[${\bf D}$]{} (35,15)[$\mbox{\boldmath$u$}^{\ast}$]{} Consider how the density $P_{(Z_1,Z_2)}$ the representative point t)$ the probability density the point to at $(Z_1,Z_2)$ at time $t$ after is emiitted from S, and to be absorbed on B after time $t+\Delta The probability density $P_{(Z_1,Z_2)}$ is obtained by integrating over the times: $$\label{eq: 3.1a} {\rm {\rm t\ t)$$ Because the diffusion process is Markovian, we can write $p$ as a product of two $p(Z_1,Z_2,t,\Delta t)=G(Z_1,Z_2,t)p_2(Z_1,Z_2,\Delta t)$, where $G(Z_1,Z_2,t)$ is the Green’s (probability density to reach after time $t$), and $p_2(Z_1,Z_2,\Delta is probability density a released $(Z_1,Z_2)$ to be on B after time $\Delta t$. From this definition of $p_2$, we have $$\label{eq: 3.1b} \int_0^\infty {\rm t\ p_2(Z_1,Z_2,\Delta that $$\label{eq: P_{(Z_1,Z_2)}=\int_0^\infty d}t\ \.$$ Strictly speaking, be the Green’s function for reaching $t$ by a path

}$]{} (90,29)[$\mbox{\boldmath$v$}$]{} (87,15)[${\bf D}$]{} (35,15)[$\mbox{\boldmAth$u$}^{\ast}$]{} ConSider How To dEtErmiNe thE probability deNSity $p_{(Z_1,Z_2)}$ for the representativE poinT tO ReacH $(z_1,Z_2)$. let $p(Z_1,z_2,t,\Delta T)$ Be THE prObAbIliTy DEnSity fOr tHe point To be at $(Z_1,Z_2)$ at TimE $t$ After it is emiITtEd from S, and To bE absorbed on B AftEr time $T+\DEltA T$. The pRobAbiliTy densITy $P_{(Z_1,Z_2)}$ iS obtained By INtegraTIng over THE tImes: $$\Label{eq: 3.1a} P_{(Z_1,Z_2)}=\int_0^\infTY {\rM D}t\int_0^\infty {\rm d}\DElta t\ p(z_1,Z_2,T,\deLTA t)$$ BEcaUse the diffUsIon prOCess is MARkOVIAn, wE Can write $p$ as a pRoduct of two FUncTions: $p(z_1,Z_2,T,\DeLTa t)=G(Z_1,Z_2,T)p_2(Z_1,Z_2,\DElTA t)$, wHere $G(Z_1,Z_2,t)$ is tHe GrEen’s functIon (proBAbility DEnsity tO reach $(z_1,Z_2)$ aFteR timE $T$), aNd $P_2(Z_1,Z_2,\deLTa t)$ IS tHe pRObaBility deNsItY for a PartICLE ReleAseD at $(Z_1,z_2)$ to be Absorbed on B afTer Time $\dEltA t$. FroM this DefiNiTion oF $p_2$, we haVe $$\labEl{Eq: 3.1b} \int_0^\infty {\rm d}\DElta T\ p_2(Z_1,Z_2,\Delta T)=1$$ so ThAt $$\lAbEl{eq: 3.2} P_{(z_1,z_2)}=\int_0^\inFty {\Rm d}T\ G(Z_1,Z_2,t) \.$$ StRictly sPEakInG, $g(z_1,z_2,t)$ Should be the Green’s fUnCTIoN for reacHing $(Z_1,Z_2)$ AFtEr TIme $t$ by a pAtH

}$]{} (90,29)[$\mbox{\bold math$v$}$] {} (8 7,1 5)[ ${ \bfD}$] {} (35,15)[$\m b ox{\ boldmath$u$}^{\ast}$]{ } Co ns i derh ow to d etermin e t h e pr ob ab ili ty de nsity $P _{(Z_1, Z_2)}$ for th erepresentati v epoint to r eac h $(Z_1,Z_2) $.Let $p (Z _1, Z _2,t, \De lta t )$ bet he pro babilityde n sity f o r the p o i nt tobe at $(Z_1,Z_2)$ at time $t$ after it is e m ii t t edfro m S, and t obe ab s orbed o n B a f ter time $t+\Delt a t$. The p r oba bility d ens i ty $P_ {(Z_1 ,Z _ 2)} $ is obtain ed b y integra ting o v er thet imes: $ $\labe l{e q:3.1a } P _{ (Z_ 1, Z _2) } =\ int _ 0^\ infty {\ rm d }t\in t_0^ \ i n f ty { \rm d}\ Delta t\ p(Z_1,Z_2 ,t, \Del t a t )$$ B ecaus e th ediffu sion p roces sis Markovian, w e ca n write $ p$as apr oduct of two fu nct ions: $ p(Z_1,Z _ 2,t ,\ D e l ta t)=G(Z_1,Z_2,t)p_ 2( Z _ 1, Z_2,\Del ta t)$ , w he r e $G(Z_1 ,Z _2, t)$i s theGree n ’s functio n (pro b ab il ity den si ty tore ach $( Z_1,Z _ 2)$aftertime $t$ ), an d $p_2(Z_1,Z_2, \ Delta t)$ ist he p ro b abil ity density fo r ap arti cler el eas e d at$(Z_1 ,Z _ 2) $ to be absorbed onBaftertime$\Delta t$. F rom this d e f i nition o f $p _ 2$ , we have $$\la bel{e q: 3.1b} \ i nt_0^\in fty { \rm d}\D elta t\ p _ 2 (Z_1,Z_2 ,\D elt a t )=1 $ $ s o that $$\lab e l {eq: 3 .2} P_{ (Z_ 1,Z_2)} =\i nt_ 0^\ inf ty {\rm d}t \ G(Z_1, Z_ 2, t) \ .$$ Stri c tly spea ki ng, $ G(Z _1,Z_ 2 ,t)$ s hould beth eG ree n’s fun c ti o n for r ea chin g $ (Z _1,Z_ 2)$a fte r time$t$ by apat h

}$]{} (90,29)[$\mbox{\boldmath$v$}$]{}_(87,15)[${\bf D}$]{}_(35,15)[$\mbox{\boldmath$u$}^{\ast}$]{} Consider how to determine_the probability_density_$P_{(Z_1,Z_2)}$ for_the_representative point to_reach $(Z_1,Z_2)$. Let_$p(Z_1,Z_2,t,\Delta t)$ be the_probability density for_the_point to be at $(Z_1,Z_2)$ at time $t$ after it is emiitted from S,_and_to be_absorbed_on_B after time $t+\Delta t$._The probability density $P_{(Z_1,Z_2)}$ is_obtained by_integrating over the times: $$\label{eq: 3.1a} P_{(Z_1,Z_2)}=\int_0^\infty {\rm d}t\int_0^\infty_{\rm_d}\Delta t\ p(Z_1,Z_2,t,\Delta_t)$$ Because the diffusion process is Markovian, we can_write $p$ as a product of_two functions: $p(Z_1,Z_2,t,\Delta_t)=G(Z_1,Z_2,t)p_2(Z_1,Z_2,\Delta_t)$,_where $G(Z_1,Z_2,t)$ is the_Green’s function (probability density to reach_$(Z_1,Z_2)$ after time $t$), and $p_2(Z_1,Z_2,\Delta_t)$ is the probability density for a_particle released at $(Z_1,Z_2)$ to be_absorbed on B after time_$\Delta t$._From this definition of $p_2$,_we have $$\label{eq:_3.1b} \int_0^\infty {\rm_d}\Delta t\ p_2(Z_1,Z_2,\Delta_t)=1$$ so that $$\label{eq: 3.2} P_{(Z_1,Z_2)}=\int_0^\infty {\rm_d}t\ G(Z_1,Z_2,t) \.$$ Strictly_speaking, $G(Z_1,Z_2,t)$ should be the Green’s_function_for reaching $(Z_1,Z_2)$_after_time_$t$ by_a path

H-\beta+h_1)m} \racc.$$ On the other hand, it follows from decorrelation argument of Lemma \[samo\] that $$\label{inm4} \log\pb\lcr \Omega_m\rcr\; \geq - \, C\, \lpa 2^{h m_1} + 2^{m} \rpa\; \geq\; -\, C \, 2^m.$$ Using (\[inm3\]) and (\[inm4\]), we can now finish the proof exactly as in the case $p < +\infty$. *(a) Contrary to [@LL1], the above proof does not require any entropy argument and relies only on an elementary estimate of the Schauder coefficients.* \(b) It would be quite interesting to calculate the optimal rate $$\gamma_0\; =\; \inf\lacc\gamma > 0\;\lva\; \lim_{\ee\downarrow 0}\ee^{\gamma}\log\pb\lcr ||M|| \leq \ee \rcr\; = \; 0\racc.\right.$$ We did not study this question in details but it seems plausible that $\gamma_0 = 0$: indeed, $M$ is a $\CC^{\infty}$ process on $(0, +\infty)$, and one may expect subexponential rate for its small deviation probabilities. Examples: particular semi-norms {#survey} =============================== In this section, we place all our results in the context of the previous literature, and show what is new. We try to be as exhaustive as possible as far as RLP’s and LFSM’s are concerned. We refer to the surveys [@LS] [@Lf] for more information about other processes and for further references. Everywhere $R$ (resp. $X$) will be an RLP (resp. an LFSM) with Hurst parameter $H > 0$ and stability index $\alpha\in (0,2]$, whereas $\|.\|$ will be a semi-norm in $\NN(\beta, p)$ or $\Nde(\beta, p)$ for some $\beta\in\R$ and $p\in (0, +\infty]$. We always suppose that $H > \beta + 1/p$ and set $$\gamma = (H-\beta-1/p)^{-1}$$ for our small deviation rate. $\k$ will stand for the small deviation constant appearing

H-\beta+h_1)m } \racc.$$ On the other hand, it follows from decorrelation argument of Lemma \[samo\ ] that $ $ \label{inm4 } \log\pb\lcr \Omega_m\rcr\; \geq - \, C\, \lpa 2^{h m_1 } + 2^{m } \rpa\; \geq\; -\, C \, 2^m.$$ use (\[inm3\ ]) and (\[inm4\ ]), we can nowadays finish the proof exactly as in the lawsuit $ p < + \infty$. * (a) Contrary to [ @LL1 ], the above validation does not require any entropy argument and trust merely on an elementary estimate of the Schauder coefficient. * \(b) It would be quite interesting to calculate the optimal rate $ $ \gamma_0\; = \; \inf\lacc\gamma > 0\;\lva\; \lim_{\ee\downarrow 0}\ee^{\gamma}\log\pb\lcr ||M|| \leq \ee \rcr\; = \; 0\racc.\right.$$ We did not study this doubt in details but it seems plausible that $ \gamma_0 = 0 $: indeed, $ M$ is a $ \CC^{\infty}$ process on $ (0, + \infty)$, and one may expect subexponential rate for its minor deviation probabilities. Examples: finical semi - norms { # survey } = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = In this section, we place all our consequence in the context of the previous literature, and show what is new. We try to be as exhaustive as potential as far as RLP ’s and LFSM ’s are concerned. We refer to the surveys [ @LS ] [ @Lf ] for more information about other process and for further reference. Everywhere $ R$ (resp. $ X$) will be an RLP (resp. an LFSM) with Hurst parameter $ H > 0 $ and stability index $ \alpha\in (0,2]$, whereas $ \|.\|$ will be a semi - norm in $ \NN(\beta, p)$ or $ \Nde(\beta, p)$ for some $ \beta\in\R$ and $ p\in (0, + \infty]$. We always suppose that $ H > \beta + 1 / p$ and sic $ $ \gamma = (H-\beta-1 / p)^{-1}$$ for our small deviation pace. $ \k$ will digest for the humble deviation constant appearing

H-\beha+h_1)m} \racc.$$ On the other hxnd, it follows yeom derorrelafion argjment of Lemma \[samo\] that $$\labxl{inn4} \log\pv\lcr \Omega_m\rcr\; \geq - \, C\, \upa 2^{h m_1} + 2^{m} \rpa\; \gwq\; -\, R \, 2^m.$$ Using (\[inm3\]) ais (\[inm4\]), wc can kow fnnmsh the proof ewactly as it the case $p < +\hnwtv$. *(a) Contrary to [@LL1], the above proof doqs not teeuire any entrjpy swgumsnt and relies only on an elementady estikate of the Svhauder coefficients.* \(b) It wluld be quite interestlng to calcomatq the optimal rate $$\gamma_0\; =\; \inf\lacc\gajma > 0\;\lva\; \lim_{\ee\downarrow 0}\ee^{\gamma}\lug\pb\ler ||M|| \leq \ee \rxr\; = \; 0\racc.\right.$$ Xe did not study tmps quesdion in details but ib seeks plausible that $\gamma_0 = 0$: indeed, $M$ is a $\CC^{\ynfty}$ proweas on $(0, +\infty)$, and inw may expact rybebpoheitizl ratf fkr its smaml deviatiob probabilities. Examklef: particular ssmi-noris {#survey} =============================== In this section, we place all our reaults in the context of the previous literatore, and shjw what is new. We try to be as exhaustive as posshble es fax as RUP’s and LFSM’s are concerned. We refer to the survqgs [@LX] [@Lf] for more lnformation about ptjet processes ana for yhrfher references. Evegywhere $R$ (rewp. $X$) will be sn RLP (resp. an LFSM) with Huest parameteg $H > 0$ and stability inbex $\alpha\in (0,2]$, wheteas $\|.\|$ eill be a semi-norm in $\NU(\beta, l)$ or $\Nde(\betw, p)$ for skoe $\beta\in\R$ and $p\kn (0, +\itfty]$. We always suppose thae $H > \beta + 1/p$ cnd set $$\eamms = (H-\beea-1/p)^{-1}$$ for oug small deviation rate. $\k$ will sdand for tje small deviation constant appxering

H-\beta+h_1)m} \racc.$$ On the other hand, it decorrelation of Lemma that $$\label{inm4} \log\pb\lcr \lpa m_1} + 2^{m} \geq\; -\, C 2^m.$$ Using (\[inm3\]) and (\[inm4\]), we now finish the proof exactly as in the case $p < +\infty$. *(a) to [@LL1], the above proof does not require any entropy argument and relies on elementary of Schauder coefficients.* \(b) It would be quite interesting to calculate the optimal rate $$\gamma_0\; =\; \inf\lacc\gamma 0\;\lva\; \lim_{\ee\downarrow 0}\ee^{\gamma}\log\pb\lcr ||M|| \leq \ee \rcr\; = 0\racc.\right.$$ We did not this question in details but seems that $\gamma_0 0$: $M$ a $\CC^{\infty}$ process $(0, +\infty)$, and one may expect subexponential rate for its small deviation probabilities. Examples: particular semi-norms {#survey} In this place all results the of the previous show what is new. We try exhaustive as possible as far as RLP’s and are concerned. refer to the surveys [@LS] [@Lf] more information about other processes and for further Everywhere $R$ (resp. $X$) will be an RLP (resp. an LFSM) with Hurst parameter $H and stability index $\alpha\in whereas $\|.\|$ will a in p)$ $\Nde(\beta, p)$ some $\beta\in\R$ and $p\in (0, +\infty]$. We always suppose that $H \beta + 1/p$ and set $$\gamma = (H-\beta-1/p)^{-1}$$ for our rate. will stand for small deviation constant appearing

H-\beta+h_1)m} \racc.$$ On the other hand, It follows fRom deCorRelAtIon aRgumEnt of Lemma \[samo\] THat $$\lAbel{inm4} \log\pb\lcr \Omega_m\rCr\; \geq - \, c\, \lPA 2^{h m_1} + 2^{m} \RPa\; \Geq\; -\, C \, 2^m.$$ using (\[inM3\]) AnD (\[INm4\]), wE cAn Now FiNIsH the pRooF exactlY as in the caSe $p < +\InFty$. *(a) Contrary TO [@Ll1], the above pRooF does not requIre Any entRoPy aRGumenT anD reliEs only ON an eleMentary esTiMAte of tHE SchaudER CoEffiCients.* \(b) It would be qUItE Interesting to cAlculaTe THe OPTimAl rAte $$\gamma_0\; =\; \inF\lAcc\gaMMa > 0\;\lva\; \liM_{\Ee\DOWNarROw 0}\ee^{\gamma}\log\pB\lcr ||M|| \leq \ee \rCR\; = \; 0\raCc.\righT.$$ WE diD Not stuDy thiS qUEstIon in detailS but It seems plAusiblE That $\gamMA_0 = 0$: indeed, $m$ is a $\CC^{\InfTy}$ pRoceSS oN $(0, +\iNftY)$, aND onE MaY exPEct SubexponEnTiAl ratE for ITS SMall DevIatiOn proBabilities. ExaMplEs: paRTicUlar sEmi-noRms {#sUrVey} =============================== In This seCtion, We Place all our resuLts iN the conteXt oF tHe pReVious LIteratUre, And Show whaT is new. WE Try To BE AS eXhaustive as possiblE aS FAr As RLP’s anD LFSM’s ARe CoNCerned. We ReFer To thE SUrveyS [@LS] [@LF] FoR more infOrmatiON aBoUt other PrOcesseS aNd fOr fUrtheR RefeRences. everywheRe $R$ (reSP. $X$) will be an RLP (rESp. an LFSM) with HURsT PArAMeteR $H > 0$ aNd stability IndeX $\AlphA\in (0,2]$, wHErEas $\|.\|$ WIll be A semi-NoRM iN $\nN(\beta, p)$ or $\Nde(\beta, p)$ foR sOme $\betA\in\R$ aNd $p\in (0, +\infty]$. We aLways suppoSE THat $H > \beta + 1/P$ and SEt $$\GAmma = (H-\beta-1/p)^{-1}$$ for oUr smaLl deviatioN Rate. $\k$ wilL stanD for the sMall deviaTIOn constaNt aPpeAriNg

H-\beta+h_1)m} \racc.$$ On the other hand , i t f ol lows fro m decorrelatio n arg ument of Lemma \[samo\ ] tha t$ $\la b el {inm4 } \log\ p b\ l c r \ Om eg a_m \r c r\ ; \ge q - \, C\, \lpa 2^{ h m _1 } + 2^{m} \r p a\ ; \geq\; - \,C \, 2^m.$$Usi ng (\[ in m3\ ] ) and (\ [inm4 \]), w e can n ow finish t h e proo f exactl y as inthe case $p < +\i n ft y $. *(a) Contr ary to [ @ LL 1 ] , t heabove proo fdoesn ot requ i re a n y e n tropy argumen t and relie s on ly onan el e mentar y est im a teof the Scha uder coeffici ents.* \(b) I t wouldbe qui teint eres t in gtoca l cul a te th e op timal ra te $ $\gam ma_0 \ ; = \; \ inf \lac c\gam ma > 0\;\lva\ ; \ lim_ { \ee \down arrow 0}\ ee ^{\ga mma}\l og\pb \l cr ||M|| \leq \ ee \ rcr\; = \ ; 0 \r acc .\ right . $$ Wedid no t study this q u est io n i ndetails but it see ms p la usible t hat $\ g am ma _ 0 = 0$:in dee d, $ M $ is a $\C C ^{ \infty}$ proce s son $(0, + \i nfty)$ ,and on e may expe ct sub exponent ial r a te for its sma l l deviation p r ob a b il i ties . Examples: p arti c ular sem i -n orm s {#su rvey} = = == = =================== == ===== In t his section,we place a l l our resu ltsi nt he context ofthe p revious li t erature, andshow wha t is new. W e try to be as ex hau s t iv e as possible a s fa ras RLP’ s a nd LFSM ’sare co nce rn ed. We re fer to t he s ur ve ys[@LS] [@Lf] fo rmor einf ormat i on abo ut ot herpr oc e sse s and f o rf u rthe rre fere nce s. Eve rywh e re$R$ (re sp. $X$)wil l bean R LP (res p. an LFSM) w it h Hurst pa ra met er $H> 0$ and s tability index $\alpha\ i n (0,2] $,where as $ \|.\|$ wi llbe a s emi - norm i n $\NN (\bet a, p) $ or $\ N d e( \be ta , p)$ fors o me$\bet a\ in\R $ and $ p\in (0, +\infty]$ . We always suppo sethat $ H> \ b et a +1/ p $ a n d set $$\gamma = (H-\beta- 1/ p )^ {-1}$$ for our s mall de viation rate . $\k$ w ill stand for thesm alld e via tion const ant appe aring

H-\beta+h_1)m} \racc.$$_On the_other hand, it follows_from decorrelation_argument_of Lemma_\[samo\]_that $$\label{inm4} \log\pb\lcr \Omega_m\rcr\;_\geq - \, C\,_\lpa 2^{h m_1}_+ 2^{m} \rpa\;_\geq\;_-\, C \, 2^m.$$ Using (\[inm3\]) and (\[inm4\]), we can now finish the proof_exactly_as in_the_case_$p < +\infty$. *(a) Contrary to_[@LL1], the above proof does_not require_any entropy argument and relies only on an_elementary_estimate of the_Schauder coefficients.* \(b) It would be quite interesting to calculate_the optimal rate $$\gamma_0\; =\; \inf\lacc\gamma_> 0\;\lva\; \lim_{\ee\downarrow 0}\ee^{\gamma}\log\pb\lcr_||M||_\leq_\ee \rcr\; = \;_0\racc.\right.$$ We did not study this_question in details but it seems_plausible that $\gamma_0 = 0$: indeed, $M$_is a $\CC^{\infty}$ process on $(0,_+\infty)$, and one may expect_subexponential rate_for its small deviation probabilities. Examples:_particular semi-norms {#survey} =============================== In_this section,_we place all_our results in the context of_the previous literature,_and show what is new. We_try_to be as_exhaustive_as_possible as_far as RLP’s_and_LFSM’s are_concerned._We refer to the surveys [@LS]_[@Lf]_for more information about other processes and_for further references. Everywhere $R$_(resp._$X$) will be an_RLP (resp. an LFSM) with_Hurst parameter $H > 0$ and_stability index_$\alpha\in (0,2]$,_whereas $\|.\|$ will be a semi-norm in $\NN(\beta, p)$ or $\Nde(\beta,_p)$ for some $\beta\in\R$ and $p\in_(0, +\infty]$. We always_suppose that_$H_> \beta +_1/p$_and set_$$\gamma = (H-\beta-1/p)^{-1}$$ for our small deviation_rate. $\k$_will stand for the small deviation_constant appearing

neighbourhood of 1 until time $2t$. As observed in [@BBS08 Lemma 4.2], there exists $C>0$ such that ${\mathbb{P}}({\mathcal{A}}_{t,\epsilon}^c) \leq C t \epsilon^{-\alpha}$. We first prove that $\lim_{t \to 0} \frac{U(t)}{t} = 1$ and $\lim_{t \to 0} \frac{R(t)}{t} = C_\alpha$ a.s. Let $\epsilon<1/2$, observe that on the event ${\mathcal{A}}_{t,\epsilon}$, we have $T > 2t$, therefore for any $s \leq t$, we have $$U(s) = \inf\left\{ r \geq 0 : \int_0^r \frac{du}{Y(u)} \geq s \right\} \in \left[ \tfrac{s}{1+\epsilon}, \tfrac{s}{1-\epsilon}\right].$$ In particular, letting $t \to 0$ we obtain $$\frac{1}{1+\epsilon} \leq \liminf_{s \to 0} \frac{U(s)}{s} \leq \limsup_{s \to 0} \frac{U(s)}{s} \leq \frac{1}{1-\epsilon} \quad \text{a.s.}$$ Letting $\epsilon \to 0$, this yields $\lim_{t \to 0} \frac{U(s)}{s} = 1$ a.s. Similarly, by we have $$\frac{1}{(1+\epsilon)^{1+\alpha}} \leq \liminf_{s \to 0} \frac{R(s)}{C_\alpha s} \leq \limsup_{s \to 0} \frac{R(s)}{C_\alpha s} \leq \frac{1}{(1-\epsilon)^{1+\alpha}},$$ yielding $\lim_{t \to 0} \frac{R(t)}{t} = C_\alpha$ a.s. We set $\tilde{R}(t) = R(t) - C_\alpha t$, we have $$\tilde{R}(t) = C_\alpha \int_0^{U(t)} \left(Y(s)^{-\alpha} - \frac{1}{Y(s)}\right) ds = C_\alpha \int_0^{U(t

neighbourhood of 1 until time $ 2t$. As observed in [ @BBS08 Lemma 4.2 ], there exists $ C>0 $ such that $ { \mathbb{P}}({\mathcal{A}}_{t,\epsilon}^c) \leq C metric ton \epsilon^{-\alpha}$. We foremost prove that $ \lim_{t \to 0 } \frac{U(t)}{t } = 1 $ and $ \lim_{t \to 0 } \frac{R(t)}{t } = C_\alpha$ a.s. Let $ \epsilon<1/2 $, note that on the consequence $ { \mathcal{A}}_{t,\epsilon}$, we have $ T > 2t$, therefore for any $ randomness \leq t$, we have $ $ U(s) = \inf\left\ { r \geq 0: \int_0^r \frac{du}{Y(u) } \geq s \right\ } \in \left [ \tfrac{s}{1+\epsilon }, \tfrac{s}{1-\epsilon}\right].$$ In especial, letting $ t \to 0 $ we obtain $ $ \frac{1}{1+\epsilon } \leq \liminf_{s \to 0 } \frac{U(s)}{s } \leq \limsup_{s \to 0 } \frac{U(s)}{s } \leq \frac{1}{1-\epsilon } \quad \text{a.s.}$$ get $ \epsilon \to 0 $, this yields $ \lim_{t \to 0 } \frac{U(s)}{s } = 1 $ a.s. Similarly, by we have $ $ \frac{1}{(1+\epsilon)^{1+\alpha } } \leq \liminf_{s \to 0 } \frac{R(s)}{C_\alpha s } \leq \limsup_{s \to 0 } \frac{R(s)}{C_\alpha s } \leq \frac{1}{(1-\epsilon)^{1+\alpha}},$$ yielding $ \lim_{t \to 0 } \frac{R(t)}{t } = C_\alpha$ a.s. We fructify $ \tilde{R}(t) = R(t) - C_\alpha t$, we accept $ $ \tilde{R}(t) = C_\alpha \int_0^{U(t) } \left(Y(s)^{-\alpha } - \frac{1}{Y(s)}\right) ds = C_\alpha \int_0^{U(t

nelghbourhood of 1 until tioe $2t$. As observeb in [@BBV08 Lemmz 4.2], there exists $C>0$ such that ${\mathbb{P}}({\methcql{A}}_{t,\ekfilon}^c) \leq C t \epsklon^{-\alpha}$. Ae first prote that $\lim_{t \to 0} \frac{U(t)}{t} = 1$ and $\pim_{t \vo 0} \frac{R(t)}{t} = C_\akpha$ a.s. Led $\epsilon<1/2$, obsesvd chat on the event ${\mathcal{A}}_{t,\epsilon}$, wq have $Y > 2t$, therefore fjr amr $s \mvq t$, we have $$U(s) = \inf\left\{ r \geq 0 : \int_0^r \hrac{du}{Y(u)} \geq s \tight\} \in \left[ \tfrac{s}{1+\epsiloj}, \tfgac{s}{1-\epsilon}\right].$$ Ij particulat, leering $t \to 0$ wd obtain $$\fgcc{1}{1+\epsilon} \lgq \liminf_{s \to 0} \frac{U(s)}{s} \leq \limsuo_{s \to 0} \frac{U(s)}{s} \oew \fgdc{1}{1-\epsilon} \qnad \teqt{a.s.}$$ Letting $\cisilon \do 0$, thix yields $\lim_{t \bo 0} \fcac{U(w)}{s} = 1$ a.s. Similarly, by xe have $$\frac{1}{(1+\epsilon)^{1+\ajpha}} \leq \nijinf_{s \to 0} \frac{R(s)}{C_\qlpha s} \leq \limryp_{s \to 0} \hrad{R(s)}{C_\aloha s} \leq \frad{1}{(1-\epsilon)^{1+\alpya}},$$ yielding $\lim_{t \to 0} \sgsc{R(t)}{t} = C_\alphz$ a.s. We sqt $\tilde{R}(t) = R(t) - C_\alpha t$, we have $$\tilde{R}(u) = C_\ampha \int_0^{U(t)} \left(Y(s)^{-\alpha} - \frac{1}{Y(s)}\right) ds = C_\alkha \int_0^{U(t

neighbourhood of 1 until time $2t$. As [@BBS08 4.2], there $C>0$ such that We prove that $\lim_{t 0} \frac{U(t)}{t} = and $\lim_{t \to 0} \frac{R(t)}{t} = a.s. Let $\epsilon<1/2$, observe that on the event ${\mathcal{A}}_{t,\epsilon}$, we have $T > therefore for any $s \leq t$, we have $$U(s) = \inf\left\{ r \geq : \frac{du}{Y(u)} s \in \left[ \tfrac{s}{1+\epsilon}, \tfrac{s}{1-\epsilon}\right].$$ In particular, letting $t \to 0$ we obtain $$\frac{1}{1+\epsilon} \leq \liminf_{s \to \frac{U(s)}{s} \leq \limsup_{s \to 0} \frac{U(s)}{s} \leq \frac{1}{1-\epsilon} \text{a.s.}$$ Letting $\epsilon \to this yields $\lim_{t \to 0} = a.s. Similarly, we $$\frac{1}{(1+\epsilon)^{1+\alpha}} \liminf_{s \to 0} s} \leq \limsup_{s \to 0} \frac{R(s)}{C_\alpha s} \leq \frac{1}{(1-\epsilon)^{1+\alpha}},$$ yielding $\lim_{t \to 0} \frac{R(t)}{t} = C_\alpha$ a.s. set $\tilde{R}(t) - C_\alpha we $$\tilde{R}(t) C_\alpha \int_0^{U(t)} \left(Y(s)^{-\alpha} ds = C_\alpha \int_0^{U(t

neighbourhood of 1 until time $2t$. as observed In [@BBS08 lemMa 4.2], tHeRe exIsts $c>0$ such that ${\mathbB{p}}({\matHcal{A}}_{t,\epsilon}^c) \leq C t \epsIlon^{-\aLpHA}$. We fIRsT provE that $\liM_{T \tO 0} \FRac{u(t)}{T} = 1$ aNd $\lIm_{T \To 0} \Frac{R(T)}{t} = C_\Alpha$ a.s. let $\epsilon<1/2$, ObsErVe that on the eVEnT ${\mathcal{A}}_{t,\EpsIlon}$, we have $T > 2t$, TheRefore FoR anY $S \leq t$, We hAve $$U(s) = \Inf\lefT\{ R \geq 0 : \inT_0^r \frac{du}{Y(U)} \gEQ s \righT\} \In \left[ \tFRAc{S}{1+\epsIlon}, \tfrac{s}{1-\epsilon}\RIgHT].$$ In particular, lEtting $T \tO 0$ We OBTaiN $$\frAc{1}{1+\epsilon} \lEq \LiminF_{S \to 0} \frac{u(S)}{s} \LEQ \LimSUp_{s \to 0} \frac{U(s)}{s} \lEq \frac{1}{1-\epsilON} \quAd \text{A.s.}$$ letTIng $\epsIlon \tO 0$, tHIs yIelds $\lim_{t \to 0} \Frac{u(s)}{s} = 1$ a.s. SimiLarly, bY We have $$\fRAc{1}{(1+\epsilOn)^{1+\alphA}} \leQ \liMinf_{S \To 0} \FrAc{R(S)}{C_\ALphA S} \lEq \lIMsuP_{s \to 0} \frac{r(s)}{c_\aLpha s} \Leq \fRAC{1}{(1-\EPsilOn)^{1+\aLpha}},$$ YieldIng $\lim_{t \to 0} \frac{r(t)}{t} = c_\alpHA$ a.s. we set $\Tilde{r}(t) = R(t) - c_\aLpha t$, We have $$\Tilde{r}(t) = c_\alpha \int_0^{U(t)} \left(y(s)^{-\alPha} - \frac{1}{Y(s)}\RigHt) Ds = C_\AlPha \inT_0^{u(t

neighbourhood of 1 untiltime $2t$. As o bse rve din [ @BBS 08 Lemma 4.2], ther e exists $C>0$ such th at ${ \m a thbb { P} }({\m athcal{ A }} _ { t,\ ep si lon }^ c )\leqC t \epsil on^{-\alph a}$ . We first pr o ve that $\li m_{ t \to 0} \fr ac{ U(t)}{ t} =1 $ and $\ lim_{ t \to0 } \fra c{R(t)}{t }= C_\al p ha$ a.s . Le t $\ epsilon<1/2$, obs e rv e that on the e vent $ {\ m at h c al{ A}} _{t,\epsil on }$, w e have $ T > 2 t $,t herefore forany $s \leq t$, we ha ve $$ U (s) =\inf\ le f t\{ r \geq 0 : \in t_0^r \fr ac{du} { Y(u)} \ g eq s \r ight\} \i n \ left [ \ tf rac {s } {1+ \ ep sil o n}, \tfrac{ s} {1 -\eps ilon } \ r i ght] .$$ Inparti cular, lettin g $ t \t o 0$ we o btain $$\ fr ac{1} {1+\ep silon }\leq \liminf_{s \to 0} \frac {U( s) }{s }\leq\ limsup _{s \t o 0} \f rac{U(s ) }{s }\ l e q\frac{1}{1-\epsilo n} \ qu ad \text {a.s.} $ $Le t ting $\e ps ilo n \t o 0$, t hisy ie lds $\li m_{t \ t o0} \frac{ U( s)}{s} = 1$ a. s. Si m ilar ly, by we have $$\f r ac{1}{(1+\epsi l on)^{1+\alpha } }\ l eq \lim inf _{s \to 0}\fra c {R(s )}{C _ \a lph a s} \ leq \ li m su p _{s \to 0} \frac{R( s) }{C_\a lphas} \leq \frac {1}{(1-\ep s i l on)^{1+\ alph a }} , $$ yielding $\ lim_{ t \to 0} \ f rac{R(t) }{t}= C_\alp ha$ a.s.W e set $\ til de{ R}( t)= R( t) - C_\alpha t $, w ehave $$ \ti lde{R}( t)= C _\a lph a\int_0^{U (t)} \le ft (Y (s )^ {-\ alpha } - \frac {1 }{Y (s )}\ right ) ds = C_\a lpha \ in t _0^ {U(t

neighbourhood_of 1_until time $2t$. As_observed in_[@BBS08_Lemma 4.2],_there_exists $C>0$ such_that ${\mathbb{P}}({\mathcal{A}}_{t,\epsilon}^c) \leq_C t \epsilon^{-\alpha}$. We first_prove that $\lim_{t_\to_0} \frac{U(t)}{t} = 1$ and $\lim_{t \to 0} \frac{R(t)}{t} = C_\alpha$ a.s. Let $\epsilon<1/2$,_observe_that on_the_event_${\mathcal{A}}_{t,\epsilon}$, we have $T >_2t$, therefore for any $s_\leq t$,_we have $$U(s) = \inf\left\{ r \geq 0_:_\int_0^r \frac{du}{Y(u)} \geq_s \right\} \in \left[ \tfrac{s}{1+\epsilon}, \tfrac{s}{1-\epsilon}\right].$$ In particular, letting_$t \to 0$ we obtain $$\frac{1}{1+\epsilon}_\leq \liminf_{s \to_0}_\frac{U(s)}{s}_\leq \limsup_{s \to 0}_\frac{U(s)}{s} \leq \frac{1}{1-\epsilon} \quad \text{a.s.}$$ Letting_$\epsilon \to 0$, this yields $\lim_{t_\to 0} \frac{U(s)}{s} = 1$ a.s. Similarly,_by we have $$\frac{1}{(1+\epsilon)^{1+\alpha}} \leq \liminf_{s_\to 0} \frac{R(s)}{C_\alpha s} \leq_\limsup_{s \to_0} \frac{R(s)}{C_\alpha s} \leq \frac{1}{(1-\epsilon)^{1+\alpha}},$$_yielding $\lim_{t \to_0} \frac{R(t)}{t}_= C_\alpha$ a.s. We_set $\tilde{R}(t) = R(t) - C_\alpha_t$, we have_$$\tilde{R}(t) = C_\alpha \int_0^{U(t)} \left(Y(s)^{-\alpha} -_\frac{1}{Y(s)}\right)_ds =_C_\alpha_\int_0^{U(t

of the model sensitivity to this parameter. The stability of the employed method can be evaluated by testing the stability to a low level change. The results for two other low levels, 132 and 154, are given in Figure \[fig: intensite\_132\_154\]. The two last curves seem to be smoother but this is partly due to the bandwidth parameter which is now equal to 0.045 (always selected by cross validation). For more precision, see Figure \[fig: comp3lambda\] where the three instantaneous intensity curves are plotted together. One can see that there is no significant difference between the three curves and that the method is rather stable. ![Similar to Figure \[fig: intensite\_111\] with a low level of 132 (left) and 154 (right)[]{data-label="fig: intensite_132_154"}](Fig7.jpg "fig:"){width="6.5cm" height="5cm"}![Similar to Figure \[fig: intensite\_111\] with a low level of 132 (left) and 154 (right)[]{data-label="fig: intensite_132_154"}](Fig8.jpg "fig:"){width="6.5cm" height="5cm"} ![Instantaneous intensity (years$^{-1}$) of the storms of level 400 obtained by extrapolation from the low levels 111 (plain line), 132 (dotted line) and 154 (dashed line)[]{data-label="fig: comp3lambda"}](Fig9.jpg){width="8cm" height="6.5cm"} A model extension ----------------- In an alternative approach, we consider the gradient of a storm to characterize its strength (instead of the ap index level). Gradients are calculated on one time step (3H) and the storm gradient is defined as the maximal gradient attained during a storm. This approach has been setting up because of the observation of storms with low levels (less than an ap index of 111) but strong effects due to fast variations of the ap index. We have led the same study with this new definition for the storm strength. The extreme gradient level are those greater than 100 and the low one is 35. The estimation of $\beta$ gives $$\hat\beta = 0.0053499\ \textrm{ with the confidence interval } \ [ 0.0038128 ; 0.006887]$$ These values are similar to those obtained with the ap index. The estimated intensity for the storms of extreme gradient is plotted in Figure

of the model sensitivity to this parameter. The stability of the employed method acting can be measure by testing the stability to a depleted grade change. The results for two early low levels, 132 and 154, are given in Figure \[fig: intensite\_132\_154\ ]. The two concluding curves seem to be fluent but this is partly due to the bandwidth argument which is now equal to 0.045 (always selected by cross validation). For more precision, attend Figure \[fig: comp3lambda\ ] where the three instantaneous intensity curves are plotted in concert. One can see that there is no meaning difference between the three curve and that the method is rather stable. ! [ Similar to Figure \[fig: intensite\_111\ ] with a humble level of 132 (left) and 154 (right)[]{data - label="fig: intensite_132_154"}](Fig7.jpg " fig:"){width="6.5 cm " height="5cm"}![Similar to Figure \[fig: intensite\_111\ ] with a low horizontal surface of 132 (leave) and 154 (right)[]{data - label="fig: intensite_132_154"}](Fig8.jpg " fig:"){width="6.5 cm " height="5 cm " } ! [ Instantaneous intensity (years$^{-1}$) of the storms of level 400 obtained by extrapolation from the low levels 111 (plain channel), 132 (dotted line) and 154 (crash line)[]{data - label="fig: comp3lambda"}](Fig9.jpg){width="8 curium " height="6.5 cm " } A model extension ----------------- In an alternate approach, we see the gradient of a storm to characterize its strength (alternatively of the ap index level). Gradients are calculated on one time step (3H) and the storm gradient is defined as the maximal gradient attained during a storm. This approach has been setting up because of the observation of storm with low level (less than an ap index of 111) but strong effects due to fast variations of the ap index. We have head the same study with this new definition for the storm forte. The extreme gradient level are those greater than 100 and the low one is 35. The estimation of $ \beta$ give $ $ \hat\beta = 0.0053499\ \textrm { with the confidence time interval } \ [ 0.0038128 ; 0.006887]$$ These values are similar to those obtained with the ap index. The estimated volume for the storms of extreme gradient is plot in Figure

of the model sensitivity tu this parametet. Rhe stebility of the dmployed method can be evalueted by ttfting the stabilith to a loa level xhanje. The results for two obker lka lereos, 132 and 154, are glven in Figgre \[fig: intenshtd\_132\_154\]. Che two last curves seem to be smootrer but tjis is partly que uo ehe gandwidth parameter which is now esual to 0.045 (always selevted by cross validation). Flr mlre precision, see Vigure \[fig: xomp3jqmbda\] where ghe three pustantaneoua intensity curves are plotted gogetker. One can swe hvat there iw no fignificant differenwe betwren the three gurvev abd that the method is rather stable. ![Simijar to Ficuxe \[fig: intensite\_111\] with a low neven of 132 (lewt) znv 154 (dight)[]{dwta-mabel="fig: ihtensite_132_154"}](Fig7.hpg "fig:"){width="6.5cm" heignt="5sn"}![Similar to Fjgure \[sid: intensite\_111\] with a low level of 132 (left) atd 154 (right)[]{data-label="fig: intebsite_132_154"}](Fig8.jpg "fig:"){width="6.5cl" height="5ci"} ![Instantaneous intensity (years$^{-1}$) of the storms of lavel 400 ubtclked ct fxtrapolation from the low levels 111 (plain line), 132 (dptned line) and 154 (dasmed line)[]{data-label="fog: cpip3lambda"}](Fig9.jpe){width="8ej" geight="6.5cm"} A model exhension ----------------- Yn an alternatyve spproach, we consider the grqdient of a frorm to characterive its strenyth (inxtead of the ap index level). Yradiehts are calfulated oh one time step (3H) anc dhe storm gradient is defiged as thx maxnmal graaienj attaiged during a storm. This approach hws begn setding up befause of the observation of stocks with low keeelv (less tkan an ap index of 111) but strong efyects duz to fxst variatpons of tie ap index. Re have led tvg same study xith this new defunition wor the storm xtrength. Nht extreme tradient level are thuae greater than 100 abd the low one os 35. Thq vstmmatijt of $\beta$ giees $$\fat\crta = 0.0053499\ \textrm{ wibh ghe vonfidence interval } \ [ 0.0038128 ; 0.006887]$$ These values arr fimilar jo those jbtained with the ap index. The tstimaved invensitu fjr the storms of extreme gradisnt is pllttcd in Figure

of the model sensitivity to this parameter. of employed method be evaluated by low change. The results two other low 132 and 154, are given in \[fig: intensite\_132\_154\]. The two last curves seem to be smoother but this is due to the bandwidth parameter which is now equal to 0.045 (always selected cross For precision, Figure \[fig: comp3lambda\] where the three instantaneous intensity curves are plotted together. One can see that is no significant difference between the three curves that the method is stable. ![Similar to Figure \[fig: with low level 132 and (right)[]{data-label="fig: intensite_132_154"}](Fig7.jpg "fig:"){width="6.5cm" to Figure \[fig: intensite\_111\] with a low level of 132 (left) and 154 (right)[]{data-label="fig: intensite_132_154"}](Fig8.jpg "fig:"){width="6.5cm" height="5cm"} intensity (years$^{-1}$) storms of 400 by from the low (plain line), 132 (dotted line) and comp3lambda"}](Fig9.jpg){width="8cm" height="6.5cm"} A model extension ----------------- In an approach, we the gradient of a storm to its strength (instead of the ap index level). are calculated on one time step (3H) and the storm gradient is defined as the attained during a storm. approach has been up of observation storms with levels (less than an ap index of 111) but strong effects to fast variations of the ap index. We have led study this new definition the storm strength. The gradient are those greater than the one estimation $\beta$ $$\hat\beta = 0.0053499\ \textrm{ the confidence interval } \ 0.0038128 ; 0.006887]$$ These obtained with the ap index. The estimated intensity the storms of extreme gradient is plotted Figure

of the model sensitivity to thIs parameteR. The sTabIliTy Of thE empLoyed method can BE evaLuated by testing the stabIlity To A Low lEVeL chanGe. The reSUlTS For TwO oTheR lOW lEvels, 132 And 154, Are giveN in Figure \[fIg: iNtEnsite\_132\_154\]. The two LAsT curves seeM to Be smoother buT thIs is paRtLy dUE to thE baNdwidTh paraMEter whIch is now eQuAL to 0.045 (alwAYs selecTED bY croSs validation). For moRE pREcision, see FiguRe \[fig: cOmP3LaMBDa\] wHerE the three iNsTantaNEous intENsITY CurVEs are plotted tOgether. One cAN seE that tHeRe iS No signIficaNt DIffErence betweEn thE three curVes and THat the mEThod is rAther sTabLe. ![SImilAR tO FIguRe \[FIg: iNTeNsiTE\_111\] wiTh a low leVeL oF 132 (left) And 154 (rIGHT)[]{Data-LabEl="fiG: inteNsite_132_154"}](Fig7.jpg "fiG:"){wiDth="6.5cM" HeiGht="5cm"}![similAr to fiGure \[fIg: inteNsite\_111\] WiTh a low level of 132 (leFt) anD 154 (right)[]{datA-laBeL="fiG: iNtensITe_132_154"}](Fig8.jPg "fIg:"){wIdth="6.5cm" hEight="5cm"} ![iNstAnTANEoUs intensity (years$^{-1}$) of ThE SToRms of levEl 400 obtaINeD bY ExtrapolAtIon From THE low lEvelS 111 (PlAin line), 132 (dOtted lINe) AnD 154 (dashed LiNe)[]{data-LaBel="Fig: Comp3lAMbda"}](fig9.jpg){Width="8cm" hEight="6.5CM"} A model extensiON ----------------- In an alternatIVe APPrOAch, wE coNsider the grAdieNT of a StorM To ChaRActerIze itS sTReNGth (instead of the ap inDeX level). gradiEnts are calculAted on one tIME Step (3H) and The sTOrM Gradient is defiNed as The maximal GRadient aTtainEd during A storm. ThiS APproach hAs bEen SetTinG UP bEcause of the obSERvatIoN of storMs wIth low lEveLs (lEss ThaN aN ap index oF 111) but stroNg EfFeCtS duE to faST variatiOnS of ThE ap Index. wE have lEd the Same StUdY WitH this neW DeFINitiOn FoR the StoRm StrenGth. THE exTreme grAdient levEl aRE thoSe GrEater thAn 100 and the low onE iS 35. The estimaTiOn oF $\beta$ gIVEs $$\hat\betA = 0.0053499\ \textrm{ with the confidencE IntervaL } \ [ 0.0038128 ; 0.006887]$$ ThEse vaLues Are similaR to Those oBtaINed witH the ap Index. thE esTIMated INTeNsiTy For the storMS Of eXtremE gRadiEnt is plOtted in Figure

of the model sensitivityto this pa ramet er. Th estab ilit y of the emplo y ed m ethod can be evaluated by t es t ingt he stab ility t o a l owle ve l c ha n ge . The re sults f or two oth erlo w levels, 13 2 a nd 154, ar e g iven in Figu re\[fig: i nte n site\ _13 2\_15 4\]. T h e twolast curv es seem t o be smo o t he r bu t this is partlyd ue to the bandwid th par am e te r whi chis now equ al to 0 . 045 (al w ay s s ele c ted by crossvalidation) . Fo r more p rec i sion,see F ig u re\[fig: comp 3lam bda\] whe re the three i n stantan eous i nte nsi ty c u rv es ar ep lot t ed to g eth er. Oneca nsee t hatt h e r e is no sig nific ant differenc e b etwe e n t he th ree c urve sand t hat th e met ho d is rather sta ble. ![Simil arto Fi gu re \[ f ig: in ten sit e\_111\ ] witha lo wl e v el of 132 (left) and 1 5 4 ( right)[] {data- l ab el = "fig: in te nsi te_1 3 2 _154" }](F i g7 .jpg "fi g:"){w i dt h= "6.5cm" h eight= "5 cm" }![ Simil a r to Figur e \[fig: inte n site\_111\] wi t h a low level of 1 32 (lef t)and 154 (ri ght) [ ]{da ta-l a be l=" f ig: i ntens it e _1 3 2_154"}](Fig8.jpg " fi g:"){w idth= "6.5cm" heigh t="5cm"} ! [ I nstantan eous in t ensity (years$ ^{-1} $) of thes torms of leve l 400 ob tained by e xtrapola tio n f rom th e lo w levels 111( p lain l ine), 1 32(dotted li ne) an d 1 54 (dashedline)[]{ da ta -l ab el= "fig: comp3lam bd a"} ]( Fig 9.jpg ) {width ="8cm " he ig ht = "6. 5cm"} A m o d el e xt en sion -- -- ----- ---- - --- In an alternat ive appr oa ch , we co nsider the gr ad ient of ast orm to ch a r acterize its strength (insteado f the a p i ndexleve l). Gradi ent s arecal c ulated on on e tim este p (3H)a n dthe s torm gradi e n t i s def in ed a s the m aximal gradient at t ain ed during a s tor m. T h i sapp r oa c h h as bee n setting up beca use of the o b se rvation of sto rm s withlow lev els ( l ess tha n an ap i ndex of 1 11 ) bu t str ong effect s due to fast var i ation s o f the ap index .Wehaveled th e sa me st udy wi th thisnew d ef initionfor the storm strength. The e xtrem e g radient l eve l ar e those g reat er than 10 0 a ndthe l owo ne is 35. Th e e s timat iono f $\beta$ gi ves $ $\ hat\beta =0 . 005 3499\ \t e xtrm{with the confidence i n terval } \ [ 0 .003 8 1 28 ;0 .006 88 7]$$ These val ues a r e similar t o those obt ained wi th the a p inde x. The estima t e di ntensi ty f orthe storm s o fe xtremegr ad i ent is plo tt ed inFigure

of_the model_sensitivity to this parameter._The stability_of_the employed_method_can be evaluated_by testing the_stability to a low_level change. The_results_for two other low levels, 132 and 154, are given in Figure \[fig: intensite\_132\_154\]._The_two last_curves_seem_to be smoother but this_is partly due to the_bandwidth parameter_which is now equal to 0.045 (always selected_by_cross validation). For_more precision, see Figure \[fig: comp3lambda\] where the three_instantaneous intensity curves are plotted together._One can see_that_there_is no significant difference_between the three curves and that_the method is rather stable. ![Similar to_Figure \[fig: intensite\_111\] with a low level_of 132 (left) and 154 (right)[]{data-label="fig:_intensite_132_154"}](Fig7.jpg "fig:"){width="6.5cm" height="5cm"}![Similar to Figure_\[fig: intensite\_111\]_with a low level of_132 (left) and_154 (right)[]{data-label="fig:_intensite_132_154"}](Fig8.jpg "fig:"){width="6.5cm" height="5cm"} ![Instantaneous_intensity (years$^{-1}$) of the storms of_level 400 obtained_by extrapolation from the low levels_111_(plain line), 132_(dotted_line)_and 154_(dashed line)[]{data-label="fig: comp3lambda"}](Fig9.jpg){width="8cm"_height="6.5cm"} A_model extension ----------------- In_an_alternative approach, we consider the gradient_of_a storm to characterize its strength (instead_of the ap index_level)._Gradients are calculated on_one time step (3H) and_the storm gradient is defined as_the maximal_gradient attained_during a storm. This approach has been setting up because of_the observation of storms with low_levels (less than an_ap index_of_111) but strong_effects_due to_fast variations of the ap index. We_have led_the same study with this new_definition for the storm_strength._The extreme gradient level are those_greater than 100 and the low_one is 35. The estimation_of_$\beta$_gives $$\hat\beta = 0.0053499\_\textrm{ with the confidence interval }_\ [ 0.0038128_ ; 0.006887]$$ These values are similar_to_those obtained with the ap index._The_estimated intensity for the storms of_extreme_gradient_is plotted in Figure

_n = w_n+w_n c_n \ge n$, $n\in\mathbb{N}$. Moreover, the monotonicity of $(c_n)_{n=1}^\infty$ and Assumption \[assumption on weight for generation\](ii) imply that $$\sum_{n=1}^{j-1} \widetilde{w}_n b_{n,j} = \sum\limits_{n=1}^{j-1} (1+c_n)w_n b_{n,j} \le (1+c_j)\sum\limits_{n=1}^{j-1} w_n b_{n,j} \le \kappa (1+c_j)w_j = \kappa \widetilde{w}_j$$ for all $j \in \mathbb{N}$. This means that Assumption \[assumption on weight for generation\] also holds for the weight $\widetilde w$. Therefore we obtain from and that $\mathcal{D}(A^{(\widetilde w)}) \subseteq \mathcal{D}(B^{(\widetilde w)})$ and $\phi_{\widetilde w}((A^{(\widetilde w)}+B^{(\widetilde w)})f) \le 0$ for $f \in \mathcal{D}(A^{(\widetilde w)})_+$, and so (f) is also satisfied. That (g) holds follows from $$\Vert A^{(w)}f \Vert_w = \sum_{n=1}^\infty w_n a_n|f_n| \le \sum_{n=1}^\infty w_n c_n|f_n| \le \sum_{n=1}^\infty \widetilde{w}_n|f_n| = \|f\|_{\widetilde w}$$ for $f \in \mathcal{D}(\tilde{A}^{(w)})_+$. Thus, the conditions of Proposition \[corollary of G=closure\] are all satisfied and therefore $G^{(w)}=\overline{A^{(w)}+B^{(w)}}$ is the generator of a substochastic $C_0$-semigroup, $(S^{(w)}(t))_{t \ge 0}$, on $\ell_w^1$, which also leaves $\mathcal{D}(C^{(w)})=\ell_{\widetilde w}^1$ invariant. Finally, assume that is satisfied and $w_n=n$ for all $

_ n = w_n+w_n c_n \ge n$, $ n\in\mathbb{N}$. Moreover, the monotonicity of $ (c_n)_{n=1}^\infty$ and Assumption   \[assumption on weight for generation\](ii) imply that $ $ \sum_{n=1}^{j-1 } \widetilde{w}_n b_{n, j } = \sum\limits_{n=1}^{j-1 } (1+c_n)w_n b_{n, joule } \le (1+c_j)\sum\limits_{n=1}^{j-1 } w_n b_{n, joule } \le \kappa (1+c_j)w_j = \kappa \widetilde{w}_j$$ for all $ joule \in \mathbb{N}$. This means that Assumption   \[assumption on weight for generation\ ] besides holds for the weight $ \widetilde w$. Therefore we receive from and that $ \mathcal{D}(A^{(\widetilde w) }) \subseteq \mathcal{D}(B^{(\widetilde w)})$ and $ \phi_{\widetilde w}((A^{(\widetilde w)}+B^{(\widetilde w)})f) \le 0 $ for $ f \in \mathcal{D}(A^{(\widetilde w)})_+$, and so (f) is also quenched. That (deoxyguanosine monophosphate) holds keep up from $ $ \Vert A^{(w)}f \Vert_w = \sum_{n=1}^\infty w_n a_n|f_n| \le \sum_{n=1}^\infty w_n c_n|f_n| \le \sum_{n=1}^\infty \widetilde{w}_n|f_n| = \|f\|_{\widetilde w}$$ for $ f \in \mathcal{D}(\tilde{A}^{(w)})_+$. Thus, the condition of Proposition   \[corollary of G = closure\ ] are all satisfied and therefore $ G^{(w)}=\overline{A^{(w)}+B^{(w)}}$ is the generator of a substochastic $ C_0$-semigroup, $ (S^{(w)}(t))_{t \ge 0}$, on $ \ell_w^1 $, which also leaves $ \mathcal{D}(C^{(w)})=\ell_{\widetilde w}^1 $ invariant. ultimately, assume that is satisfied and $ w_n = n$ for all $

_n = a_n+w_n c_n \ge n$, $n\in\mathbb{N}$. Ooreover, the mouitonicmty of $(d_n)_{n=1}^\infty$ and Assumption \[assumption on wwight for generation\](ii) implh that $$\sul_{n=1}^{j-1} \widerildt{w}_n b_{n,j} = \sum\limiva_{n=1}^{j-1} (1+c_n)w_k b_{n,j} \le (1+e_j)\wum\limits_{n=1}^{j-1} w_n b_{n,j} \le \kdppa (1+c_j)w_j = \ka[px \cidetilde{w}_j$$ for all $j \in \mathbb{N}$. Thif means tjat Assumption \[wssuk[tioh on weight for generation\] also homds for the weight $\wodetilde w$. Therefore we obhain from and that $\matjcal{D}(A^{(\widetulde q)}) \subseteq \mxthcal{D}(B^{(\widetilde w)})$ ans $\phi_{\widetilde w}((A^{(\widetilde w)}+B^{(\wiaetilbe w)})f) \le 0$ fir $f \lt \mathcal{D}(A^{(\xidetijde w)})_+$, and so (f) is alvo satixfied. That (g) hpldv fillows from $$\Vert A^{(w)}f \Tert_w = \sum_{n=1}^\infty w_n w_n|f_n| \le \vuj_{n=1}^\infty w_n c_n|f_n| \oe \sum_{n=1}^\hnftf \wiawtiude{s}_n|h_n| = \|f\|_{\widftimde w}$$ for $r \in \mathcao{D}(\tilde{A}^{(w)})_+$. Thus, the cpnqpyions of Prolositijn \[sorollary of G=closure\] are all satisfied ans therefore $G^{(w)}=\overline{A^{(q)}+B^{(w)}}$ is the generator lf a subseochastic $C_0$-semigroup, $(S^{(w)}(t))_{t \ge 0}$, on $\ell_w^1$, which also neavea $\mauhgwu{E}(C^{(a)})=\ell_{\widetilde w}^1$ invariant. Finally, assume that ya xanisfied and $w_n=n$ fjr all $

_n = w_n+w_n c_n \ge n$, $n\in\mathbb{N}$. monotonicity $(c_n)_{n=1}^\infty$ and \[assumption on weight \widetilde{w}_n = \sum\limits_{n=1}^{j-1} (1+c_n)w_n \le (1+c_j)\sum\limits_{n=1}^{j-1} w_n \le \kappa (1+c_j)w_j = \kappa \widetilde{w}_j$$ all $j \in \mathbb{N}$. This means that Assumption \[assumption on weight for generation\] holds for the weight $\widetilde w$. Therefore we obtain from and that $\mathcal{D}(A^{(\widetilde \subseteq w)})$ $\phi_{\widetilde w)}+B^{(\widetilde w)})f) \le 0$ for $f \in \mathcal{D}(A^{(\widetilde w)})_+$, and so (f) is also satisfied. That holds follows from $$\Vert A^{(w)}f \Vert_w = \sum_{n=1}^\infty a_n|f_n| \le \sum_{n=1}^\infty w_n \le \sum_{n=1}^\infty \widetilde{w}_n|f_n| = \|f\|_{\widetilde for \in \mathcal{D}(\tilde{A}^{(w)})_+$. the of \[corollary of G=closure\] all satisfied and therefore $G^{(w)}=\overline{A^{(w)}+B^{(w)}}$ is the generator of a substochastic $C_0$-semigroup, $(S^{(w)}(t))_{t \ge 0}$, on $\ell_w^1$, also leaves invariant. Finally, that satisfied $w_n=n$ for all

_n = w_n+w_n c_n \ge n$, $n\in\mathbb{N}$. MoreoVer, the monoTonicIty Of $(c_N)_{n=1}^\InftY$ and assumption \[assuMPtioN on weight for generation\](Ii) impLy THat $$\sUM_{n=1}^{J-1} \wideTilde{w}_n B_{N,j} = \SUM\liMiTs_{N=1}^{j-1} (1+c_N)w_N B_{n,J} \le (1+c_j)\Sum\Limits_{n=1}^{J-1} w_n b_{n,j} \le \kaPpa (1+C_j)W_j = \kappa \widetILdE{w}_j$$ for all $j \In \mAthbb{N}$. This meAns That AsSuMptIOn \[assUmpTion oN weighT For genEration\] alSo HOlds foR The weigHT $\WiDetiLde w$. Therefore we obTAiN From and that $\matHcal{D}(A^{(\WiDEtILDe w)}) \SubSeteq \mathcAl{d}(B^{(\widETilde w)})$ aND $\pHI_{\WIdeTIlde w}((A^{(\widetilDe w)}+B^{(\widetilDE w)})f) \Le 0$ for $f \In \MatHCal{D}(A^{(\wIdetiLdE W)})_+$, anD so (f) is also sAtisFied. That (g) Holds fOLlows frOM $$\Vert A^{(w)}F \Vert_w = \Sum_{N=1}^\inFty w_N A_n|F_n| \Le \sUm_{N=1}^\InfTY w_N c_n|F_N| \le \Sum_{n=1}^\inftY \wIdEtildE{w}_n|f_N| = \|F\|_{\WIDetiLde W}$$ for $F \in \maThcal{D}(\tilde{A}^{(w)})_+$. thuS, the COndItionS of PrOposItIon \[coRollarY of G=cLoSure\] are all satisFied And therefOre $g^{(w)}=\OveRlIne{A^{(w)}+b^{(W)}}$ is the GenEraTor of a sUbstochAStiC $C_0$-SEMIgRoup, $(S^{(w)}(t))_{t \ge 0}$, on $\ell_w^1$, whIcH ALsO leaves $\mAthcal{d}(c^{(w)})=\ElL_{\WidetildE w}^1$ InvAriaNT. finalLy, asSUmE that is sAtisfiED aNd $W_n=n$ for aLl $

_n = w_n+w_n c_n \ge n$, $ n\in\mathb b{N}$ . M ore ov er,themonotonicity o f $(c _n)_{n=1}^\infty$ andAssum pt i on \ [ as sumpt ion onw ei g h t f or g ene ra t io n\](i i)imply t hat $$\sum _{n =1 }^{j-1} \wid e ti lde{w}_n b _{n ,j} = \sum \li mits_{ n= 1}^ { j-1}(1+ c_n)w _n b_{ n ,j} \le (1+c_ j) \ sum\li m its_{n= 1 } ^{ j-1} w_n b_{n,j} \l e \ k appa (1+c_j)w_ j =\k a pp a \wi det ilde{w}_j$ $for a l l $j \i n \ m a t hbb { N}$. This mea ns that Ass u mpt ion \[ as sum p tion o n wei gh t fo r generatio n\]also hold s fort he weig h t $\wid etilde w$ . T here f or eweob t ain fr oma ndthat $\m at hc al{D} (A^{ ( \ w i deti lde w)} ) \su bseteq \mathc al{ D}(B ^ {(\ widet ildew)}) $and $ \phi_{ \wide ti lde w}((A^{(\wi deti lde w)}+B ^{( \w ide ti lde w ) })f) \ le0$for $f\in \ma t hca l{ D } ( A^ {(\widetilde w)})_ +$ , an d so (f) is al s osa t isfied.Th at(g)h o lds f ollo w sfrom $$\ Vert A ^ {( w) }f \Ver t_ w = \s um _{n =1} ^\inf t y w_ n a_n| f_n| \ le \s u m_{n=1}^\infty w_n c_n|f_n| \ l e \ s um_{ n=1 }^\infty \w idet i lde{ w}_n | f_ n| = \| f\|_{ \w i de t ilde w}$$ for $f \i n\mathc al{D} (\tilde{A}^{( w)})_+$. T h u s, the c ondi t io n s of Propositi on \[ corollaryo f G=clos ure\] are all satisfie d and ther efo re$G^ {(w ) } =\ overline{A^{( w ) }+B^ {( w)}}$ i s t he gene rat orofa s ub stochasti c $C_0$- se mi gr ou p,$(S^{ ( w)}(t))_ {t \g e0}$ , on$ \ell_w ^1$,whic hal s o l eaves $ \ ma t h cal{ D} (C ^{(w )}) =\ ell_{ \wid e til de w}^1 $ invaria nt. Fin al ly , assum e that is sat is fied and $ w_ n=n $ fora l l $

_n =_w_n+w_n c_n_\ge n$, $n\in\mathbb{N}$. Moreover,_the monotonicity_of_$(c_n)_{n=1}^\infty$ and_Assumption \[assumption_on weight for_generation\](ii) imply that_$$\sum_{n=1}^{j-1} \widetilde{w}_n b_{n,j} _= \sum\limits_{n=1}^{j-1} (1+c_n)w_n_b_{n,j} _ \le (1+c_j)\sum\limits_{n=1}^{j-1} w_n b_{n,j} \le \kappa (1+c_j)w_j = \kappa \widetilde{w}_j$$ for_all_$j \in_\mathbb{N}$._This_means that Assumption \[assumption on weight_for generation\] also holds for_the weight_$\widetilde w$. Therefore we obtain from and that_$\mathcal{D}(A^{(\widetilde_w)}) \subseteq \mathcal{D}(B^{(\widetilde_w)})$ and $\phi_{\widetilde w}((A^{(\widetilde w)}+B^{(\widetilde w)})f) \le 0$ for_$f \in \mathcal{D}(A^{(\widetilde w)})_+$, and so_(f) is also_satisfied._That_(g) holds follows from_$$\Vert A^{(w)}f \Vert_w = \sum_{n=1}^\infty w_n_a_n|f_n| \le \sum_{n=1}^\infty w_n c_n|f_n| _ \le \sum_{n=1}^\infty \widetilde{w}_n|f_n| = \|f\|_{\widetilde_w}$$ for $f \in \mathcal{D}(\tilde{A}^{(w)})_+$. Thus, the_conditions of Proposition \[corollary of G=closure\]_are all_satisfied and therefore $G^{(w)}=\overline{A^{(w)}+B^{(w)}}$ is_the generator of_a substochastic_$C_0$-semigroup, $(S^{(w)}(t))_{t \ge_0}$, on $\ell_w^1$, which also leaves_$\mathcal{D}(C^{(w)})=\ell_{\widetilde w}^1$ invariant. Finally,_assume that is satisfied and $w_n=n$_for_all $

to the nuclear spin environment via the contact hyperfine interaction, although they can still undergo decay due to spin-orbit coupling. The decay may still occur on an even longer time scale than for electrons,[@bulaev:2005b] which suggests the dot-confined hole spin may be another good candidate for quantum computing. Alternatively, quantum dots fabricated in isotopically purified $^{28}\mathrm{Si}$[@friesen:2003a] or $^{12}\mathrm{C}$ nanotubes[@mason:2004a; @sapmaz:2006a; @graeber:2006a] would be free of nuclei with spin, and therefore free of hyperfine-induced decoherence. While the field of quantum-dot spin decoherence has been very active in the last few years, there still remain significant misconceptions regarding the nature of the most relevant (hyperfine) coupling, particularly, the range of validity of semiclassical spin models and traditional decoherence methods involving ensemble averaging have been called into question for a single isolated quantum dot with a potentially controllable environment. We address these issues in section \[sec:hyperfine\]. Hyperfine interaction: quantum and classical dynamics {#sec:hyperfine} ===================================================== Exponential decay of the longitudinal and transverse components of spin is typically measured by the decay time scales $T_1$ and $T_2$, respectively.[@slichter:1980a] The longitudinal spin relaxation rate $1/T_1$ due to spin-orbit interaction and phonon emission is significantly reduced in quantum dots relative to the bulk in the presence of a weak Zeeman splitting $B=|\mathbf{B}|$ and large orbital level spacing $\hbar\omega_0$ ($1/T_1\propto B^5/(\hbar\omega_0)^4$).[@khaetskii:2001a; @golovach:2004a] This decay time has been shown to be on the order of $T_1\sim1\,\mathrm{ms}$ in gated GaAs quantum dots at $B\approx8\,\mathrm{T}$,[@elzerman:2004a] and to reach a value as large as $T_1=170\,\mathrm{ms}$ at low magnetic fields ($B=1.75\,\mathrm{T}$).[@amasha:2006a] Furthermore, since dephasing is absent at leading order for fluctuations that couple through the spin-orbit interaction, the $T_2$

to the nuclear spin environment via the contact hyperfine interaction, although they can still undergo decay due to spin - orbit yoke. The decay may however occur on an even longer meter plate than for electrons,[@bulaev:2005b ] which suggests the dot - confined fix spin may be another good campaigner for quantum computer science. Alternatively, quantum dots fabricate in isotopically purified $ ^{28}\mathrm{Si}$[@friesen:2003a ] or $ ^{12}\mathrm{C}$ nanotubes[@mason:2004a; @sapmaz:2006a; @graeber:2006a ] would be free of lens nucleus with tailspin, and therefore free of hyperfine - induce decoherence. While the field of quantum - dot tailspin decoherence has been very active in the last few years, there still stay significant misconceptions see the nature of the most relevant (hyperfine) yoke, particularly, the stove of validity of semiclassical spin models and traditional decoherence method involving ensemble averaging have been called into question for a single isolated quantum dot with a potentially controllable environment. We address these issues in section \[sec: hyperfine\ ]. Hyperfine interaction: quantum and classical dynamics { # sec: hyperfine } = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = Exponential decay of the longitudinal and transverse components of spin is typically measured by the decay time scales $ T_1 $ and $ T_2 $, respectively.[@slichter:1980a ] The longitudinal tailspin easiness rate $ 1 / T_1 $ due to tailspin - eye socket interaction and phonon discharge is significantly reduced in quantum dot relative to the bulk in the presence of a weak Zeeman burst $ B=|\mathbf{B}|$ and large orbital level spacing $ \hbar\omega_0 $ ($ 1 / T_1\propto B^5/(\hbar\omega_0)^4$).[@khaetskii:2001a; @golovach:2004a ] This decay time has been testify to be on the order of $ T_1\sim1\,\mathrm{ms}$ in gated GaAs quantum dots at $ B\approx8\,\mathrm{T}$,[@elzerman:2004a ] and to reach a value as large as $ T_1=170\,\mathrm{ms}$ at gloomy magnetic fields ($ B=1.75\,\mathrm{T}$).[@amasha:2006a ] Furthermore, since dephasing is absent at leading order for fluctuation that couple through the spin - orbit interaction, the $ T_2 $

to the nuclear spin envirokment via the couract hbperfins interaztion, although they can stilp ynderto decay due to spin-orcit couplpng. The dwcay nay still occur on an evsk lonyec time scale thsn for elewtrons,[@bulaev:2005b] fhkck suggests the dot-confined hole spin may be ajother good cagdidsee fkg wuantum computing. Alternativemy, quanuum dots fabricatec in isotopically purified $^{28}\matjrm{Si}$[@friesen:2003a] or $^{12}\mwthrm{C}$ nanojhbef[@nason:2004a; @sapmax:2006a; @graeber:2006a] would be ftee of nuclei with spin, and therdfore free of hi'wrflte-induced dxcohervnce. While the field ox quantim-dot spin decphecencw has been very activx in the last few yewrs, there scill remain significabt miscmnce[tiovw rdgaddmng the nwtuce of the mkst relevanr (hyperfine) couplinb, [qrticularly, tge rande of validity of semiclassical spin modens znd traditional decoherwnce methods involvinh ensemblq averaging have been called into question for a vinglx ksooabed dyajtum dot with a potentially controllable enviwknkekt. We address thcse issues in sectooj \[xgc:hyperfine\]. Hypgrfine nhtsraction: quantum ajd clasfical dynamics {#sec:nyperfine} ===================================================== Exponential decay if the longinudibal and transverse components of xpin os typically measured bv the secay time dcales $T_1$ zvd $T_2$, respectivelh.[@slpchtar:1980a] The longitudinal spin welaxatioi ratz $1/T_1$ due go skin-orbie interactlon akg phonon emission ls siynifiwantly redkced in quantum dots relative to the bulk in jhe prvsence of a wesk Zeeman splytting $B=|\mathbf{Y}|$ and laxge orcital levem spaciig $\hbar\omega_0$ ($1/T_1\propto B^5/(\hbas\lmega_0)^4$).[@khaetskmi:2001a; @golovwch:2004a] Thiw decay gime has been xhown to yt on the oeder of $T_1\sim1\,\mathrm{ks}$ kh gated GaAs qucuton dots at $B\apptox8\,\oatrrl{T}$,[@xlzeridn:2004a] and to raach a xslue xs largt aw $T_1=170\,\mayhrm{ms}$ at low magnethc fjelds ($B=1.75\,\mathrm{T}$).[@amasna:2006w] Furthetmore, sinse dephasing os absent at leadijg orver foc flucyuajions that couple through the slin-orbit lntcraction, the $E_2$

to the nuclear spin environment via the interaction, they can undergo decay due may occur on an longer time scale for electrons,[@bulaev:2005b] which suggests the dot-confined spin may be another good candidate for quantum computing. Alternatively, quantum dots fabricated isotopically purified $^{28}\mathrm{Si}$[@friesen:2003a] or $^{12}\mathrm{C}$ nanotubes[@mason:2004a; @sapmaz:2006a; @graeber:2006a] would be free of nuclei spin, therefore of decoherence. While the field of quantum-dot spin decoherence has been very active in the last few there still remain significant misconceptions regarding the nature the most relevant (hyperfine) particularly, the range of validity semiclassical models and decoherence involving averaging have been into question for a single isolated quantum dot with a potentially controllable environment. We address these issues section \[sec:hyperfine\]. quantum and dynamics ===================================================== decay of the transverse components of spin is typically decay time scales $T_1$ and $T_2$, respectively.[@slichter:1980a] The spin relaxation $1/T_1$ due to spin-orbit interaction and emission is significantly reduced in quantum dots relative the bulk in the presence of a weak Zeeman splitting $B=|\mathbf{B}|$ and large orbital level ($1/T_1\propto B^5/(\hbar\omega_0)^4$).[@khaetskii:2001a; @golovach:2004a] This time has been to on order $T_1\sim1\,\mathrm{ms}$ in GaAs quantum dots at $B\approx8\,\mathrm{T}$,[@elzerman:2004a] and to reach a value as as $T_1=170\,\mathrm{ms}$ at low magnetic fields ($B=1.75\,\mathrm{T}$).[@amasha:2006a] Furthermore, since dephasing at order for fluctuations couple through the spin-orbit the

to the nuclear spin environmeNt via the coNtact HypErfInE intEracTion, although thEY can Still undergo decay due to Spin-oRbIT couPLiNg. The Decay maY StILL ocCuR oN an EvEN lOnger TimE scale tHan for elecTroNs,[@Bulaev:2005b] which SUgGests the doT-coNfined hole spIn mAy be anOtHer GOod caNdiDate fOr quanTUm compUting. AlteRnATively, QUantum dOTS fAbriCated in isotopicalLY pURified $^{28}\mathrm{Si}$[@FrieseN:2003a] OR $^{12}\mATHrm{c}$ naNotubes[@masOn:2004A; @sapmAZ:2006a; @graebER:2006a] WOULd bE Free of nuclei wIth spin, and tHEreFore frEe Of hYPerfinE-induCeD DecOherence. WhiLe thE field of qUantum-DOt spin dECoherenCe has bEen VerY actIVe In The LaST feW YeArs, THerE still reMaIn SigniFicaNT MISconCepTionS regaRding the naturE of The mOSt rElevaNt (hypErfiNe) CouplIng, parTiculArLy, the range of valIditY of semiclAssIcAl sPiN modeLS and trAdiTioNal decoHerence MEthOdS INVoLving ensemble averaGiNG HaVe been caLled inTO qUeSTion for a SiNglE isoLATed quAntuM DoT with a poTentiaLLy CoNtrollaBlE envirOnMenT. We AddreSS theSe issuEs in sectIon \[seC:Hyperfine\]. HyperFIne interactioN: QuANTuM And cLasSical dynamiCs {#seC:HypeRfinE} ===================================================== exPonENtial Decay Of THe LOngitudinal and transVeRse comPonenTs of spin is typIcally measURED by the deCay tIMe SCales $T_1$ and $T_2$, respEctivEly.[@slichteR:1980A] The longItudiNal spin rElaxation RATe $1/T_1$ due to SpiN-orBit IntERAcTion and phonon EMIssiOn Is signiFicAntly reDucEd iN quAntUm Dots relatIve to the BuLk In ThE prEsencE Of a weak ZEeMan SpLitTing $B=|\MAthbf{B}|$ And laRge oRbItAL leVel spacINg $\HBAr\omEgA_0$ ($1/T_1\PropTo B^5/(\HbAr\omeGa_0)^4$).[@khAEtsKii:2001a; @golOvach:2004a] ThiS deCAy tiMe HaS been shOwn to be on the oRdEr of $T_1\sim1\,\maThRm{mS}$ in gatED gaAs quanTum dots at $B\approx8\,\mathrm{T}$,[@ELzerman:2004A] anD to reAch a Value as laRge As $T_1=170\,\matHrm{MS}$ at low MagnetIc fieLdS ($B=1.75\,\mATHrm{T}$).[@aMAShA:2006a] FUrThermore, siNCE dePhasiNg Is abSent at lEading order for flucTUatIons that couplE thRougH THe SpiN-OrBIt iNtERacTIOn, the $T_2$

to the nuclear spin envir onment via thecon tac thype rfin e interaction, alth ough they can still un dergo d e cayd ue to s pin-orb i tc o upl in g. Th ed ec ay ma y s till oc cur on aneve nlonger times ca le than fo r e lectrons,[@b ula ev:200 5b ] w h ich s ugg eststhe do t -confi ned holesp i n mayb e anoth e r g oodcandidate for qua n tu m computing. Al ternat iv e ly , qua ntu m dots fab ri cated in isot o pi c a l lyp urified $^{28 }\mathrm{Si } $[@ friese n: 200 3 a] or$^{12 }\ m ath rm{C}$ nano tube s[@mason: 2004a; @sapmaz : 2006a;@graeb er: 200 6a]w ou ld be f r eeo fnuc l eiwith spi n, a nd th eref o r e free of hyp erfin e-induced dec ohe renc e . While thefiel dof qu antum- dot s pi n decoherence h as b een veryact iv e i nthe l a st few ye ars , there stillr ema in s i gn ificant misconcept io n s r egarding the n a tu re of the m os t r elev a n t (hy perf i ne ) coupli ng, pa r ti cu larly,th e rang eofval idity of s emicla ssical s pin m o dels and tradi t ional decoher e nc e me t hods in volving ens embl e ave ragi n ghav e been call ed in t o question for a si ng le iso lated quantum dotwith a pot e n t ially co ntro l la b le environment . Weaddress th e se issue s insection\[sec:hyp e r fine\]. Hy per fin e i n t er action: quant u m and c lassica l d ynamics {# sec :hy per fi ne} ===== ======== == == == == === ===== = ======== == === == === ===== Expon entia l de ca yo f t he long i tu d i nalan dtran sve rs e com pone n tsof spin is typic all y mea su re d by th e decay timesc ales $T_1$ a nd$T_2$, r espectiv ely.[@slichter:1980a] T h e longi tud inalspin relaxati onrate $ 1/T _ 1$ due to sp in-or bi t i n t eract i o nand p honon emis s i onis si gn ific antly r educed in quantumd ots relative tothe bul k in th e p r ese nc e of a weak Zeeman sp litting $B =| \ ma thbf{B}|$a ndla rge orb ital le vel s p acing $ \hbar\ome ga_0$ ($1 /T _1\p r o pto B^5/(\hba r\omega_ 0)^4$).[@ k haets k ii :2001 a;@golov ac h:2 004a] Thisd eca y tim e hasbe en sho wn to b e on the order of $T_1\sim1\,\m athrm{ ms}$ingated GaA s q u ant um dots a t $B \approx8\, \ma thr m{T}$ ,[@ e lzerm an:2 0 04 a]a nd to rea c h a value as la r g eas $T_1=170 \ , \ mat hrm{m s}$ at low mag netic fields ($B= 1 .75\,\mathrm{T }$). [ @ ama sha : 2006 a] Furthermore,sin ce d ephasing i s absent at leading o r der f or flu ctuati ons tha t co u ple th roug h t he spin-o rbi ti nteract io n, the $T _2$

to_the nuclear_spin environment via the_contact hyperfine_interaction,_although they_can_still undergo decay_due to spin-orbit_coupling. The decay may_still occur on_an_even longer time scale than for electrons,[@bulaev:2005b] which suggests the dot-confined hole spin may_be_another good_candidate_for_quantum computing. Alternatively, quantum dots_fabricated in isotopically purified $^{28}\mathrm{Si}$[@friesen:2003a]_or $^{12}\mathrm{C}$_nanotubes[@mason:2004a; @sapmaz:2006a; @graeber:2006a] would be free of nuclei_with_spin, and therefore_free of hyperfine-induced decoherence. While the field of quantum-dot spin_decoherence has been very active in_the last few_years,_there_still remain significant misconceptions_regarding the nature of the most_relevant (hyperfine) coupling, particularly, the range_of validity of semiclassical spin models and_traditional decoherence methods involving ensemble averaging_have been called into question_for a_single isolated quantum dot with_a potentially controllable_environment. We_address these issues_in section \[sec:hyperfine\]. Hyperfine interaction: quantum and_classical dynamics {#sec:hyperfine} ===================================================== Exponential_decay of the longitudinal and transverse_components_of spin is_typically_measured_by the_decay time scales_$T_1$_and $T_2$,_respectively.[@slichter:1980a]_The longitudinal spin relaxation rate $1/T_1$_due_to spin-orbit interaction and phonon emission is_significantly reduced in quantum_dots_relative to the bulk_in the presence of a_weak Zeeman splitting $B=|\mathbf{B}|$ and large_orbital level_spacing $\hbar\omega_0$_($1/T_1\propto B^5/(\hbar\omega_0)^4$).[@khaetskii:2001a; @golovach:2004a] This decay time has been shown to be_on the order of $T_1\sim1\,\mathrm{ms}$ in_gated GaAs quantum dots_at $B\approx8\,\mathrm{T}$,[@elzerman:2004a]_and_to reach a_value_as large_as $T_1=170\,\mathrm{ms}$ at low magnetic fields ($B=1.75\,\mathrm{T}$).[@amasha:2006a]_Furthermore, since_dephasing is absent at leading order_for fluctuations that couple_through_the spin-orbit interaction, the $T_2$

FTU device [@Pucella] by means of the presence of electric field stochastic fluctuations. We have analyzed the dynamics of an electron subjected to an external magnetic field (which mimics the Tokamak toroidal one), and a transverse stochastic electric field reproducing the electrostatic turbulence profile, mainly due to non-linear drift phenomena. We have found the proper microscopic value of the electrostatic turbulence requested to reproduce the FTU phenomenology. The relevant stochastic perturbation is found to occur at the typical microscopic scale of the Tokamak plasma, [*i.e.,* ]{}few milliliters. As an explanation of such a phenomenological issue, we have inferred that the Debye shield on the electron loses its efficiency near the critical density value. [00]{} J. Wesson, *Tokamaks* (Clarendon Press, Oxford, 1997). G. Pucella et al., *Nucl. Fus.* **53**, 023007 (2013). M. Greenwald, *Pl. Phys. Contr. Fus.* **44**, R27 (2002). A. Carati et al., *Chaos* **22**, 033124 (2012). L. Brillouin, *Phys. Rev.* **67**, 260 (1945). D.A. Gates, L. Delgado-Aparicio, *Phys. Rev. Lett.* **43**, 267 (2012). A.H. Boozer, *Phys. Plasmas* **12**, 104502 (2005). C. Iglesias, J. Lebowitz, D. MacGowan, *Phys. Rev. A* **28**, 1667 (1983). J. Jimenez-Aquino, M. Romero-Bastida, *Phys. Rev. E* **76**, 021106 (2007). J. Wei, E. Norman, *Proc. Amer. Math. Soc.* **15**, 2034065 (1964). O. Tudisco et al., *Fus. Eng. Des.* **85**, 902 (2010). D. Bohm, *The characteristics of electrical discharges in magnetic fields*, Eds. A. Guthrie and R.K. Wakerling (MacGraw-Hill, New York, 1949). L. Spitzer, *Phys. Fluids* **3**(4), 6

FTU device [ @Pucella ] by means of the presence of electric field stochastic fluctuations. We have analyze the moral force of an electron subjected to an external magnetic sphere (which mimic the Tokamak toroidal one), and a transverse stochastic electric field reproducing the electrostatic turbulence visibility, chiefly due to non - linear drift phenomena. We have found the proper microscopic value of the electrostatic turbulence requested to reproduce the FTU phenomenology. The relevant stochastic disruption is found to occur at the distinctive microscopic scale of the Tokamak plasma, [ * i.e., * ] { } few milliliters. As an explanation of such a phenomenological issue, we have inferred that the Debye shield on the electron loses its efficiency near the critical density value. [ 00 ] { } J. Wesson, * Tokamaks * (Clarendon Press, Oxford, 1997). G. Pucella et al. , * Nucl. Fus. * * * 53 * *, 023007 (2013). M. Greenwald, * Pl. Phys. Contr. Fus. * * * 44 * *, R27 (2002). A. Carati et al. , * Chaos * * * 22 * *, 033124 (2012). L. Brillouin, * Phys. Rev. * * * 67 * *, 260 (1945). D.A. Gates, L. Delgado - Aparicio, * Phys. Rev. Lett. * * * 43 * *, 267 (2012). A.H. Boozer, * Phys. Plasmas * * * 12 * *, 104502 (2005). C. Iglesias, J. Lebowitz, D. MacGowan, * Phys. Rev. A * * * 28 * *, 1667 (1983). J. Jimenez - Aquino, M. Romero - Bastida, * Phys. Rev. E * * * 76 * *, 021106 (2007). J. Wei, E. Norman, * Proc. Amer. Math. Soc. * * * 15 * *, 2034065 (1964). O. Tudisco et al. , * Fus. Eng. Des. * * * 85 * *, 902 (2010). D. Bohm, * The feature of electrical discharges in magnetic field *, Eds. A. Guthrie and R.K. Wakerling (MacGraw - Hill, New York, 1949). L. Spitzer, * Phys. Fluids * * * 3**(4), 6

FTK device [@Pucella] by meanr of the presenew of enectrid field rtochastic fluctuations. We heve qnalyevd the dynamics of an electron subjectwd ti an externem magnebnc fispd (wkirh mimics the Tpkamak tormidal one), and d grcnsverse stochastic electric field rqproduconh the electrosjatic eurbhlence profile, mainly due to non-lihear drpft phenomena. We nave found the proper micrlscooic value of the epectrostatix tuwvulence requdsted to rtpxoduce the RTU phenomenology. The relevant rtochcstic pertutyqtilt is found vo occlr at the typlbal micsoscopiv scale of the Tonamqk plasma, [*i.e.,* ]{}few millmliters. As an explanwtion of vueh a phenomenological iwsue, fe hdve kbfefres vhaf the Febbe shield oh the electeon loses its efficoegbu near the cditicaj qensity value. [00]{} J. Wesson, *Tokamaks* (Clarendot Pdess, Oxford, 1997). G. Pucella er al., *Nucl. Fus.* **53**, 023007 (2013). M. Gregnwald, *Pl. [hys. Contr. Fus.* **44**, R27 (2002). A. Carati et al., *Chaos* **22**, 033124 (2012). L. Brillogin, *Piyr. Rtv.* **67**, 260 (1945). D.A. Tahes, L. Delgado-Aparicio, *Phys. Rev. Lett.* **43**, 267 (2012). A.H. Boozqd, *Khyx. Plasmas* **12**, 104502 (2005). C. Lglesias, J. Lebowita, F. KwcGowan, *Phys. Tev. A* **28**, 1667 (1983). J. Nimenez-Aquino, M. Rolero-Basjida, *Pyys. Rev. E* **76**, 021106 (2007). J. Wei, E. Norman, *Proc. Amer. Maty. Soc.* **15**, 2034065 (1964). O. Tubisxo et al., *Fus. Eng. Dzs.* **85**, 902 (2010). D. Bohm, *Che chsractrristics of electrical bischadges in magjetic fiemas*, Eds. A. Guthrie anc S.K. Wakerling (MacGraw-Hill, Nqw York, 1949). L. Spiczer, *Phyr. Floids* **3**(4), 6

FTU device [@Pucella] by means of the electric stochastic fluctuations. have analyzed the to external magnetic field mimics the Tokamak one), and a transverse stochastic electric reproducing the electrostatic turbulence profile, mainly due to non-linear drift phenomena. We have the proper microscopic value of the electrostatic turbulence requested to reproduce the FTU The stochastic is to occur at the typical microscopic scale of the Tokamak plasma, [*i.e.,* ]{}few milliliters. As an of such a phenomenological issue, we have inferred the Debye shield on electron loses its efficiency near critical value. [00]{} Wesson, (Clarendon Oxford, 1997). G. et al., *Nucl. Fus.* **53**, 023007 (2013). M. Greenwald, *Pl. Phys. Contr. Fus.* **44**, R27 (2002). A. et al., 033124 (2012). Brillouin, Rev.* 260 (1945). D.A. Delgado-Aparicio, *Phys. Rev. Lett.* **43**, 267 *Phys. Plasmas* **12**, 104502 (2005). C. Iglesias, J. D. MacGowan, Rev. A* **28**, 1667 (1983). J. M. Romero-Bastida, *Phys. Rev. E* **76**, 021106 (2007). Wei, E. Norman, *Proc. Amer. Math. Soc.* **15**, 2034065 (1964). O. Tudisco et al., *Fus. **85**, 902 (2010). D. *The characteristics of discharges magnetic Eds. Guthrie and Wakerling (MacGraw-Hill, New York, 1949). L. Spitzer, *Phys. Fluids* **3**(4), 6

FTU device [@Pucella] by means of The presencE of elEctRic FiEld sTochAstic fluctuatiONs. We Have analyzed the dynamicS of an ElECtroN SuBjectEd to an eXTeRNAl mAgNeTic FiELd (Which MimIcs the TOkamak toroIdaL oNe), and a transvERsE stochastiC elEctric field rEprOducinG tHe eLEctroStaTic tuRbulenCE profiLe, mainly dUe TO non-liNEar drifT PHeNomeNa. We have found the pROpER microscopic vaLue of tHe ELeCTRosTatIc turbulenCe RequeSTed to rePRoDUCE thE fTU phenomenolOgy. The relevANt sTochasTiC peRTurbatIon is FoUNd tO occur at the TypiCal microsCopic sCAle of thE tokamak Plasma, [*I.e.,* ]{}fEw mIlliLItErS. As An EXplANaTioN Of sUch a phenOmEnOlogiCal iSSUE, We haVe iNferRed thAt the Debye shiEld On thE EleCtron Loses Its eFfIcienCy near The crItIcal density valuE. [00]{} J. WeSson, *TokamAks* (clAreNdOn PreSS, OxforD, 1997). G. PUceLla et al., *nucl. Fus.* **53**, 023007 (2013). m. greEnWALD, *PL. Phys. Contr. Fus.* **44**, R27 (2002). A. CarAtI ET aL., *Chaos* **22**, 033124 (2012). L. BRillouIN, *PHyS. rev.* **67**, 260 (1945). D.A. GatEs, l. DeLgadO-aParicIo, *PhYS. REv. Lett.* **43**, 267 (2012). A.H. boozer, *pHyS. PLasmas* **12**, 104502 (2005). C. igLesias, j. LEboWitZ, D. MacgOwan, *phys. ReV. A* **28**, 1667 (1983). J. JimenEz-AquINo, M. Romero-BastiDA, *Phys. Rev. E* **76**, 021106 (2007). J. Wei, e. noRMAn, *pRoc. AMer. math. Soc.* **15**, 2034065 (1964). O. TudIsco ET al., *FUs. EnG. deS.* **85**, 902 (2010). D. BOHm, *The CharaCtERiSTics of electrical disChArges iN magnEtic fields*, Eds. a. Guthrie anD r.k. wakerlinG (MacgRaW-hill, New York, 1949). L. SpItzer, *phys. Fluids* **3**(4), 6

FTU device [@Pucella] bymeans of t he pr ese nce o f el ectr ic field stoch a stic fluctuations. We have anal yz e d th e d ynami cs of a n e l e ctr on s ubj ec t ed to a n e xternal magneticfie ld (which mimi c sthe Tokama k t oroidal one) , a nd a t ra nsv e rse s toc hasti c elec t ric fi eld repro du c ing th e electr o s ta ticturbulence profil e ,m ainly due to n on-lin ea r d r i ftphe nomena. We h ave f o und the pr o p e r m i croscopic val ue of the e l ect rostat ic tu r bulenc e req ue s ted to reprodu ce t he FTU ph enomen o logy. T h e relev ant st och ast ic p e rt ur bat io n is fo und tooccur at t he typi calm i c r osco pic sca le of the Tokamakpla sma, [*i .e.,* ]{}f ew m il lilit ers. A s anex planation of su ch a phenomen olo gi cal i ssue, we hav e i nfe rred th at theD eby es h i el d on the electronlo s e sits effi ciency ne ar the crit ic aldens i t y val ue.[0 0]{} J. Wesso n ,*T okamaks *(Clare nd onPre ss, O x ford , 1997 ). G. P ucell a et al., *Nucl . Fus.* **53** , 0 2 3 00 7 (20 13) . M. Green wald , *Pl . Ph y s. Co n tr. F us.*** 4 4* * , R27 (2002). A. C ar ati et al., *Chaos* **22 **, 033124 ( 2 012). L . Br i ll o uin, *Phys. Re v.* * *67**, 260 (1945). D.A. Gates,L. Delgad o - Aparicio , * Phy s.Rev . Le tt.* **43**,2 6 7 (2 01 2). A. H.Boozer, *P hys . P las ma s* **12** , 104502 ( 20 05 ). C . Igl e sias, J. L ebo wi tz, D. M a cGowan , *Ph ys.Re v. A***28**, 16 6 7 (19 83 ). J. Ji me nez-A quin o , M . Romer o-Bastida , * P hys. R ev . E* ** 76**, 021106(2 007). J.We i,E. Nor m a n, *Proc . Amer. Math. Soc.* **1 5 **, 203 406 5 (19 64). O. Tudi sco et al .,* Fus. E ng. De s.* * *8 5** , 902 ( 2 0 10 ). D . Bohm, *T h e ch aract er isti cs of e lectrical discharg e s i n magnetic fi eld s*,E d s. A. Gu t hri ea ndR . K. Wakerling (M acGraw-Hil l, Ne w York, 19 4 9). L. Spit zer, *P hys.F luids***3**(4), 6

FTU_device [@Pucella]_by means of the_presence of_electric_field stochastic_fluctuations._We have analyzed_the dynamics of_an electron subjected to_an external magnetic_field_(which mimics the Tokamak toroidal one), and a transverse stochastic electric field reproducing the_electrostatic_turbulence profile,_mainly_due_to non-linear drift phenomena. We_have found the proper microscopic_value of_the electrostatic turbulence requested to reproduce the FTU_phenomenology._The relevant stochastic_perturbation is found to occur at the typical microscopic_scale of the Tokamak plasma, [*i.e.,*_]{}few milliliters. As_an_explanation_of such a phenomenological_issue, we have inferred that the_Debye shield on the electron loses_its efficiency near the critical density value. [00]{} J._Wesson, *Tokamaks* (Clarendon Press, Oxford, 1997). G._Pucella et al., *Nucl. Fus.*_**53**, 023007_(2013). M. Greenwald, *Pl. Phys. Contr._Fus.* **44**, R27_(2002). A. Carati_et al., *Chaos*_**22**, 033124 (2012). L. Brillouin, *Phys. Rev.*_**67**, 260 (1945). D.A._Gates, L. Delgado-Aparicio, *Phys. Rev. Lett.*_**43**,_267 (2012). A.H. Boozer,_*Phys._Plasmas*_**12**, 104502_(2005). C. Iglesias, J._Lebowitz,_D. MacGowan,_*Phys._Rev. A* **28**, 1667 (1983). J. Jimenez-Aquino,_M._Romero-Bastida, *Phys. Rev. E* **76**, 021106 (2007). J._Wei, E. Norman, *Proc._Amer._Math. Soc.* **15**, 2034065_(1964). O. Tudisco et al., *Fus._Eng. Des.* **85**, 902 (2010). D. Bohm,_*The characteristics_of electrical_discharges in magnetic fields*, Eds. A. Guthrie and R.K. Wakerling (MacGraw-Hill,_New York, 1949). L. Spitzer, *Phys. Fluids*_**3**(4), 6

combined. By exploiting the $\Theta$-functions the result can be rewritten as $$\begin{aligned} \label{28} - \lim_{\Delta\lambda\rightarrow 0} \frac{\delta{\cal H}_1(\lambda,\Delta\lambda)}{\Delta\lambda} &=& \sum_{{\bf k},{\bf q},\sigma} \delta\left( \left| \varepsilon_{\bf k} - \varepsilon_{({\bf k}+{\bf q})} \right| + \omega_{\bf q} - \lambda \right) \frac{ \left| g_{\bf q} \right|^{2} \Theta\left[ \omega_{\bf q} - \left| \varepsilon_{\bf k} - \varepsilon_{({\bf k}+{\bf q})} \right| \right] }{ \left| \varepsilon_{\bf k} - \varepsilon_{({\bf k}+{\bf q})} \right| + \omega_{\bf q} }\times \nonumber \\[1ex] && \qquad\qquad\times c_{({\bf k}+{\bf q}),\sigma}^{\dagger} c_{-({\bf k}+{\bf q}),-\sigma}^{\dagger} c_{-{\bf k},-\sigma} c_{{\bf k},\sigma}.\end{aligned}$$ Here, we have assumed $g_{\bf q}=g_{-{\bf q}}$. Eq.  describes the renormalization of the $\lambda$ dependent Hamiltonian ${\cal H}_{\lambda}$ with respect to the cutoff $\lambda$. Next we use to derive flow equations for the parameters $\Delta_{{\bf k},\lambda}$ and $C_{\lambda}$. For this purpose, a factorization with respect to the full Hamiltonian ${\cal H}$ is carried out. The final flow equations read $$\begin{aligned} \frac{d\Delta_{{\bf k},\lambda}}{d\lambda} &=& -2 \sum_{\bf q} \delta\left( \left| \varepsilon_{\bf k} - \varepsilon_{({\bf k}+{\bf q})} \right| + \omega_{\bf q} - \lambda \right) \frac{ \left| g_{\bf q} \right|^{2}

combined. By exploiting the $ \Theta$-functions the result can be rewritten as $ $ \begin{aligned } \label{28 } - \lim_{\Delta\lambda\rightarrow 0 } \frac{\delta{\cal H}_1(\lambda,\Delta\lambda)}{\Delta\lambda } & = & \sum_{{\bf k},{\bf q},\sigma } \delta\left ( \left| \varepsilon_{\bf k } - \varepsilon_{({\bf k}+{\bf q }) } \right| + \omega_{\bf q } - \lambda \right) \frac { \left| g_{\bf q } \right|^{2 } \Theta\left [ \omega_{\bf q } - \left| \varepsilon_{\bf k } - \varepsilon_{({\bf k}+{\bf q }) } \right| \right ] } { \left| \varepsilon_{\bf k } - \varepsilon_{({\bf k}+{\bf q }) } \right| + \omega_{\bf q } } \times \nonumber \\[1ex ] & & \qquad\qquad\times c_{({\bf k}+{\bf q}),\sigma}^{\dagger } c_{-({\bf k}+{\bf q}),-\sigma}^{\dagger } c_{-{\bf k},-\sigma } c_{{\bf k},\sigma}.\end{aligned}$$ Here, we have assume $ g_{\bf q}=g_{-{\bf q}}$. Eq.   trace the renormalization of the $ \lambda$ dependent Hamiltonian $ { \cal H}_{\lambda}$ with respect to the cutoff $ \lambda$. Next we practice to derive flow equations for the parameter $ \Delta_{{\bf k},\lambda}$ and $ C_{\lambda}$. For this purpose, a factorization with respect to the entire Hamiltonian $ { \cal H}$ is carried out. The final menstruation equations read $ $ \begin{aligned } \frac{d\Delta_{{\bf k},\lambda}}{d\lambda } & = & -2 \sum_{\bf q } \delta\left ( \left| \varepsilon_{\bf k } - \varepsilon_{({\bf k}+{\bf q }) } \right| + \omega_{\bf q } - \lambda \right) \frac { \left| g_{\bf q } \right|^{2 }

colbined. By exploiting the $\Theta$-functions the revult czn be redritten as $$\begin{aligned} \labxl{28} - \lim_{\Dtjta\lambda\rightarrod 0} \frac{\dvlta{\cal H}_1(\oambva,\Delta\lambda)}{\Delvz\lambda} &=& \sum_{{\gn k},{\bf w},\sigma} \delta\keft( \lext| \varepviuou_{\bf k} - \varepsilon_{({\bf k}+{\bf q})} \right| + \omegs_{\bv q} - \lambda \ryght) \frzb{ \left| g_{\bf q} \right|^{2} \Theta\lsft[ \omega_{\bf q} - \left| \varepsilon_{\bf k} - \varfpsilon_{({\bf k}+{\bf q})} \rihht| \righj] }{ \left| \xarepsilon_{\bf k} - \varepsjlon_{({\bf k}+{\bf q})} \right| + \omega_{\bf d} }\tikes \nonumbgx \\[1ex] && \qquad\qquad\vimes c_{({\bf k}+{\bf q}),\sinka}^{\daggar} c_{-({\bg k}+{\bf q}),-\sigma}^{\danger} c_{-{\bd k},-\sigma} c_{{\bf k},\sigma}.\xnd{aligned}$$ Here, we hwve assumad $g_{\bf q}=g_{-{\bf q}}$. Eq.  dewceibes the renuemauizztmon of thf $\lembda$ depensent Hamiltinian ${\cal H}_{\lambda}$ wotr respect to tge cutjfs $\lambda$. Next we use to derive flow equdtikns for the parameters $\Eelta_{{\bf k},\lambda}$ and $C_{\pambda}$. Fow this purpose, a factorization with respect to tha fulm Hanijginlan ${\cal H}$ is carried out. The final flow equatyknx gead $$\begin{aligned} \frac{d\Delts_{{\bv l},\jambda}}{d\lambda} &=& -2 \sum_{\bf q} \delta\left( \lfft| \vaeepsilon_{\bs k} - \varepsilon_{({\bf k}+{\bf q})} \righr| + \omega_{\by q} - \lambda \right) \frac{ \leyt| g_{\bf q} \ribht|^{2}

combined. By exploiting the $\Theta$-functions the result rewritten $$\begin{aligned} \label{28} \lim_{\Delta\lambda\rightarrow 0} \frac{\delta{\cal \delta\left( \varepsilon_{\bf k} - k}+{\bf q})} \right| \omega_{\bf q} - \lambda \right) \frac{ g_{\bf q} \right|^{2} \Theta\left[ \omega_{\bf q} - \left| \varepsilon_{\bf k} - \varepsilon_{({\bf k}+{\bf \right| \right] }{ \left| \varepsilon_{\bf k} - \varepsilon_{({\bf k}+{\bf q})} \right| + \omega_{\bf }\times \\[1ex] \qquad\qquad\times k}+{\bf q}),\sigma}^{\dagger} c_{-({\bf k}+{\bf q}),-\sigma}^{\dagger} c_{-{\bf k},-\sigma} c_{{\bf k},\sigma}.\end{aligned}$$ Here, we have assumed $g_{\bf q}=g_{-{\bf q}}$. describes the renormalization of the $\lambda$ dependent Hamiltonian H}_{\lambda}$ with respect to cutoff $\lambda$. Next we use derive equations for parameters k},\lambda}$ $C_{\lambda}$. For this a factorization with respect to the full Hamiltonian ${\cal H}$ is carried out. The final flow equations $$\begin{aligned} \frac{d\Delta_{{\bf -2 \sum_{\bf \delta\left( \varepsilon_{\bf - \varepsilon_{({\bf k}+{\bf + \omega_{\bf q} - \lambda \right) q} \right|^{2}

combined. By exploiting the $\ThEta$-functioNs the ResUlt CaN be rEwriTten as $$\begin{aliGNed} \lAbel{28} - \lim_{\Delta\lambda\righTarroW 0} \fRAc{\deLTa{\Cal H}_1(\lAmbda,\DeLTa\LAMbdA)}{\DElTa\lAmBDa} &=& \Sum_{{\bf K},{\bf Q},\sigma} \dElta\left( \leFt| \vArEpsilon_{\bf k} - \vaREpSilon_{({\bf k}+{\bf Q})} \riGht| + \omega_{\bf q} - \lAmbDa \righT) \fRac{ \LEft| g_{\bF q} \rIght|^{2} \THeta\leFT[ \omega_{\Bf q} - \left| \vaRePSilon_{\bF K} - \varepsILOn_{({\Bf k}+{\bF q})} \right| \right] }{ \left| \vARePSilon_{\bf k} - \varepsIlon_{({\bf K}+{\bF Q})} \rIGHt| + \oMegA_{\bf q} }\times \nOnUmber \\[1EX] && \qquad\qQUaD\TIMes C_{({\Bf k}+{\bf q}),\sigma}^{\daGger} c_{-({\bf k}+{\bf q}),-\SIgmA}^{\daggeR} c_{-{\Bf k},-\SIgma} c_{{\bF k},\sigMa}.\ENd{aLigned}$$ Here, wE havE assumed $g_{\Bf q}=g_{-{\bf Q}}$. eq.  descrIBes the rEnormaLizAtiOn of THe $\LaMbdA$ dEPenDEnT HaMIltOnian ${\cal h}_{\lAmBda}$ wiTh reSPECT to tHe cUtofF $\lambDa$. Next we use to DerIve fLOw eQuatiOns foR the PaRametErs $\DelTa_{{\bf k},\LaMbda}$ and $C_{\lambda}$. FOr thIs purpose, A faCtOriZaTion wITh respEct To tHe full HAmiltonIAn ${\cAl h}$ IS CaRried out. The final flOw EQUaTions reaD $$\begin{ALiGnED} \frac{d\DeLtA_{{\bf K},\lamBDA}}{d\lamBda} &=& -2 \sUM_{\bF q} \delta\lEft( \lefT| \VaRePsilon_{\bF k} - \VarepsIlOn_{({\bF k}+{\bF q})} \rigHT| + \omeGa_{\bf q} - \lAmbda \rigHt) \fraC{ \Left| g_{\bf q} \right|^{2}

combined. By exploiting t he $\Theta $-fun cti ons t he r esul t can be rewri t tenas $$\begin{aligned} \lab el { 28} - \lim _{\Delt a \l a m bda \r ig hta rr o w0} \fr ac{\del ta{\cal H} _1( \l ambda,\Delta \ la mbda)}{\De lta \lambda} &= & \sum_ {{ \bf k},{\ bfq},\s igma} \del ta\left( \left | \ va reps ilon_{\bf k} - \v a re p silon_{({\bf k }+{\bf q } )} \ri ght| + \ omega _ {\bf q} -\ l a mbd a \right) \frac{ \le ft| g_ {\ bfq } \rig ht|^{ 2} \Theta\lef t[ \ome ga_{\b f q} - \l eft| \ var eps ilon _ {\ bf k} - \va r ep sil o n_{ ({\bf k} +{ \b f q}) } \r i g h t | \ righ t] }{ \left| \ var epsil on_{\ bf k }- \va repsil on_{( {\ bf k}+{\bf q})} \right|+ \ om ega _{ \bf q } }\t ime s \ nonumbe r \\[1e x ] & &\ qq uad\qquad\times c_ { ( {\ bf k}+{\ bf q}) , \s ig m a}^{\dag ge r} c _ { -({\b f k} + {\ bf q}),- \sigma } ^{ \d agger} c_{-{\ bf k} ,-\ sigma } c_{{\b f k},\si gma}. \ end{aligned}$$ Here, we have as s u me d $g_ {\b f q}=g_{-{\ bf q } }$.Eq.de scr i bes t he re no r ma l ization of the $\la mb da$ de pende nt Hamiltonia n ${\cal H } _ { \lambda} $ wi t hr espect to thecutof f $\lambda $ . Next w e use to deri ve flow e q u ations f orthe pa ram e t er s $\Delta_{{\ b f k}, \l ambda}$ an d $C_{\ lam bda }$. Fo rthis purp ose, a f ac to ri za tio n wit h respect t o t he fu ll Ha m iltoni an ${ \cal H }$ iscarried ou t . The f in al f low e quati onsr ead $$\beg in{aligne d} \fr ac {d \Delta_ {{\bf k},\lam bd a}}{d\lamb da } & =& - 2 \sum_{\b f q} \delta\left( \left| \ vare psilon_{\ bfk} - \ var e psilon _{({\b f k}+ {\ bfq } )} \r igh t| + \om e g a_{ \bf q }- \l ambda \right) \frac { \left| g_{\ bfq} \ r i gh t|^ { 2}

combined._By exploiting_the $\Theta$-functions the result_can be_rewritten_as $$\begin{aligned} __\label{28} -_\lim_{\Delta\lambda\rightarrow 0} _\frac{\delta{\cal H}_1(\lambda,\Delta\lambda)}{\Delta\lambda} &=& _\sum_{{\bf k},{\bf q},\sigma}_ _ \delta\left( \left| \varepsilon_{\bf k}_-_\varepsilon_{({\bf k}+{\bf_q})} __ \right| _ + \omega_{\bf q} -_\lambda _\right) \frac{ \left|_g_{\bf_q} \right|^{2} _ \Theta\left[ \omega_{\bf_q} - _ \left|_\varepsilon_{\bf_k}_- \varepsilon_{({\bf k}+{\bf q})}_\right| \right] _}{ \left| _ \varepsilon_{\bf k} -_\varepsilon_{({\bf k}+{\bf q})} _\right| + \omega_{\bf q} _}\times \nonumber_\\[1ex] && \qquad\qquad\times c_{({\bf_k}+{\bf q}),\sigma}^{\dagger} _ c_{-({\bf_k}+{\bf q}),-\sigma}^{\dagger} _c_{-{\bf k},-\sigma} c_{{\bf k},\sigma}.\end{aligned}$$_Here, we have_assumed $g_{\bf q}=g_{-{\bf q}}$. Eq.  describes_the_renormalization of the_$\lambda$_dependent_Hamiltonian ${\cal_H}_{\lambda}$ with respect_to_the cutoff_$\lambda$._Next we use to derive flow_equations_for the parameters $\Delta_{{\bf k},\lambda}$ and $C_{\lambda}$._For this purpose, a_factorization_with respect to the_full Hamiltonian ${\cal H}$ is_carried out. The final flow equations_read $$\begin{aligned} _ \frac{d\Delta_{{\bf_k},\lambda}}{d\lambda} &=& -2 \sum_{\bf q} \delta\left( _\left| _ \varepsilon_{\bf k} -_\varepsilon_{({\bf k}+{\bf_q})} _ _\right| _ _ + \omega_{\bf q} - \lambda _\right) _ \frac{ _\left| g_{\bf q} \right|^{2}

decreases exponentially with the period of 21 minutes $\pm$ 1.\] This half-life agrees with the presently adopted value of 22.00(7) min. This observation of $^{223}$Fr also represented the discovery of the element francium. $^{224-226}$Fr {#fr-16.unnumbered} -------------- Hansen et al. reported the first observation of $^{224}$Fr, $^{225}$Fr and $^{226}$Fr in the paper “Decay characteristics of short-lived radio-nuclides studied by on-line isotope separator techniques” in 1969 [@1969Han01]. Protons of 600 MeV from the CERN synchrocyclotron bombarded a molten tin target and $^{224}$Fr, $^{225}$Fr and $^{226}$Fr were separated using the ISOLDE facility. The paper summarized the ISOLDE program and did not contain details about the individual nuclei but the results were presented in a table. The measured half-life of 2.67(20) min for $^{224}$Fr agrees with the presently adopted value of 3.33(10) min and the 3.9(2) min for $^{225}$Fr is included in the calculation of the currently accepted half-life of 3.95(14) min. The half-life of 1.43(23) min for $^{226}$Fr is within a factor of two of the present value of 49(1) s. $^{227,228}$Fr {#fr-17.unnumbered} -------------- In 1972 Klapisch et al. reported the first observation of $^{227}$Fr and $^{228}$Fr in “Half-life of the new isotope $^{32}$Na; Observation of $^{33}$Na and other new isotopes produced in the reaction of high-energy protons on U” [@1972Kla01]. Uranium targets were bombarded with 24 GeV protons from the CERN proton synchrotron. $^{227}$Fr and $^{228}$Fr were identified by on-line mass spectrometry and decay curves were measured. “Following the same procedure as for Na, the isotopes $^{48}$K, $^{49}$K, and $^{50}$K were found. However, their half-lives were not short compared with the diffusion time, and hence could not be determined. We also observed the new neutron-rich isotopes $^{227}$Fr and $^{228}$Fr produced in the spallation of

decreases exponentially with the period of 21 minutes $ \pm$ 1.\ ] This half - life match with the soon adopted value of 22.00(7)   min. This notice of $ ^{223}$Fr also defend the discovery of the element francium. $ ^{224 - 226}$Fr { # fr-16.unnumbered } -------------- Hansen et al.   reported the inaugural observation of $ ^{224}$Fr, $ ^{225}$Fr and $ ^{226}$Fr in the paper “ Decay feature of short - lived radio receiver - nuclides studied by on - line isotope separator techniques ” in 1969 [ @1969Han01 ]. proton of 600   MeV from the CERN synchrocyclotron bombarded a molten tin prey and $ ^{224}$Fr, $ ^{225}$Fr and $ ^{226}$Fr were separated using the iseult facility. The paper summarized the iseult program and did not contain details about the individual nuclei but the results were show in a table. The measured half - life of 2.67(20)   min for $ ^{224}$Fr agrees with the soon adopted value of 3.33(10)   min and the 3.9(2)   min for $ ^{225}$Fr is included in the calculation of the currently accepted half - biography of 3.95(14)   min. The half - life of 1.43(23)   min for $ ^{226}$Fr is within a factor of two of the present value of 49(1)   s. $ ^{227,228}$Fr { # fr-17.unnumbered } -------------- In 1972 Klapisch et al.   reported the first observation of $ ^{227}$Fr and $ ^{228}$Fr in “ Half - life of the new isotope $ ^{32}$Na; Observation of $ ^{33}$Na and other fresh isotopes produced in the chemical reaction of eminent - department of energy protons on U ” [ @1972Kla01 ]. Uranium targets were bombarded with 24 GeV protons from the CERN proton synchrotron. $ ^{227}$Fr and $ ^{228}$Fr were identified by on - line mass spectrometry and decay bend were measured. “ Following the same procedure as for Na, the isotope $ ^{48}$K, $ ^{49}$K, and $ ^{50}$K were found. However, their half - lives were not short compared with the diffusion time, and hence could not be decide. We also observed the new neutron - rich isotopes $ ^{227}$Fr and $ ^{228}$Fr grow in the spallation of

defreases exponentially wiuh the period of 21 minutev $\pm$ 1.\] Fhis halw-life agrees with the presenvly qdopttb value of 22.00(7) min. This ubservatiln of $^{223}$Fr alsi representxs the dlfcovsvy of vhe element frakcium. $^{224-226}$Fr {#fr-16.utnumbered} -------------- Hanset dt al. reported the first observation os $^{224}$Fr, $^{225}$Fr ajd $^{226}$Fr in the pwper “Qecag characteristics of short-lived rasio-nuclpdes studied by om-line isotope separator tefhnieues” in 1969 [@1969Han01]. Protojs of 600 MeV ftkm eye CERN syncfrocyclotron bombarded a molten tin target and $^{224}$Fr, $^{225}$Fr xnd $^{226}$Fx were sepatcred osing the ISILDE sacility. The paper sgmmarizrd the ISOLDE irogrem abd did not contain devails about the indididual nuwlzi but the results weee presgnted in x tacle. Tie jeasurfd ialf-life of 2.67(20) min for $^{224}$Fr agrees with the prtsegnky adopted vzlue os 3.33(10) iin and the 3.9(2) min for $^{225}$Fr is included in tve dalculation of the currwntly accepted half-live of 3.95(14) min. The half-life of 1.43(23) min for $^{226}$Fr is within a factor of two kw tkc oeedent value of 49(1) s. $^{227,228}$Fr {#fr-17.unnumbered} -------------- In 1972 Klapisch et zl. teiorted the first jbservation ov $^{227}$Gt and $^{228}$Fr in “Hauf-life of the new isotope $^{32}$Nw; Obserdatiob of $^{33}$Na agd oyher new isotopes produced un the reactpon if high-energy protlns on U” [@1972Klc01]. Uranoum tsrgets were bombarded wnth 24 GsV protons vrom the DDRN proton synchfotgon. $^{227}$Xr and $^{228}$Fr were identified fy on-line masx spectfomejry and decay curges wcse measured. “Followlng tke sake proceduge as for Na, the isotopes $^{48}$K, $^{49}$K, aiv $^{50}$K were founc. Vowvver, theix half-kives were noe short compargd with tke difwusion timv, and henre could not be determineg. We also obsxrved the new neurron-ricf isotopes $^{227}$Fr amd $^{228}$Fr probbced in tye spallation of

decreases exponentially with the period of 21 1.\] half-life agrees the presently adopted observation $^{223}$Fr also represented discovery of the francium. $^{224-226}$Fr {#fr-16.unnumbered} -------------- Hansen et reported the first observation of $^{224}$Fr, $^{225}$Fr and $^{226}$Fr in the paper “Decay of short-lived radio-nuclides studied by on-line isotope separator techniques” in 1969 [@1969Han01]. Protons 600 from CERN bombarded a molten tin target and $^{224}$Fr, $^{225}$Fr and $^{226}$Fr were separated using the ISOLDE facility. paper summarized the ISOLDE program and did not details about the individual but the results were presented a The measured of min $^{224}$Fr agrees with presently adopted value of 3.33(10) min and the 3.9(2) min for $^{225}$Fr is included in the calculation the currently of 3.95(14) The of min for $^{226}$Fr a factor of two of the 49(1) s. $^{227,228}$Fr {#fr-17.unnumbered} -------------- In 1972 Klapisch al. reported first observation of $^{227}$Fr and $^{228}$Fr “Half-life of the new isotope $^{32}$Na; Observation of and other new isotopes produced in the reaction of high-energy protons on U” [@1972Kla01]. Uranium bombarded with 24 GeV from the CERN synchrotron. and were by on-line spectrometry and decay curves were measured. “Following the same procedure as Na, the isotopes $^{48}$K, $^{49}$K, and $^{50}$K were found. However, were short compared with diffusion time, and hence not determined. We also observed neutron-rich $^{227}$Fr in spallation

decreases exponentially witH the period Of 21 minUteS $\pm$ 1.\] thIs haLf-liFe agrees with thE PresEntly adopted value of 22.00(7) min. this oBsERvatIOn Of $^{223}$Fr aLso reprESeNTEd tHe DiScoVeRY oF the eLemEnt franCium. $^{224-226}$Fr {#fr-16.unNumBeRed} -------------- Hansen et aL. RePorted the fIrsT observation Of $^{224}$FR, $^{225}$Fr and $^{226}$fr In tHE papeR “DeCay chAracteRIstics Of short-liVeD Radio-nUClides sTUDiEd by On-line isotope sepaRAtOR techniques” in 1969 [@1969HAn01]. ProtOnS Of 600 mEv frOm tHe CERN syncHrOcyclOTron bomBArDED A moLTen tin target aNd $^{224}$Fr, $^{225}$Fr and $^{226}$Fr WEre SeparaTeD usINg the IsOLDE FaCIliTy. The paper sUmmaRized the IsOLDE pROgram anD Did not cOntain DetAilS aboUT tHe IndIvIDuaL NuCleI But The resulTs WeRe preSentED IN A tabLe. THe meAsureD half-life of 2.67(20) miN foR $^{224}$Fr aGReeS with The prEsenTlY adopTed valUe of 3.33(10) mIn And the 3.9(2) min for $^{225}$Fr iS incLuded in thE caLcUlaTiOn of tHE curreNtlY acCepted hAlf-life OF 3.95(14) miN. THE HAlF-life of 1.43(23) min for $^{226}$Fr is wItHIN a Factor of Two of tHE pReSEnt value Of 49(1) S. $^{227,228}$Fr {#Fr-17.unNUMbereD} -------------- In 1972 KLApIsch et al. ReportED tHe First obSeRvatioN oF $^{227}$Fr And $^{228}$fr in “HALf-liFe of thE new isotOpe $^{32}$Na; oBservation of $^{33}$Na ANd other new isoTOpES PrODuceD in The reaction Of hiGH-eneRgy pROtOns ON U” [@1972Kla01]. uraniUm TArGEts were bombarded witH 24 GEV protOns frOm the CERN protOn synchrotRON. $^{227}$fr and $^{228}$Fr wEre iDEnTIfied by on-line mAss spEctrometry ANd decay cUrves Were measUred. “FolloWINg the samE prOceDurE as FOR NA, the isotopes $^{48}$K, $^{49}$k, ANd $^{50}$K wErE found. HOweVer, theiR haLf-lIveS weRe Not short cOmpared wItH tHe DiFfuSion tIMe, and henCe CouLd Not Be detERmined. we alsO obsErVeD The New neutROn-RICh isOtOpEs $^{227}$Fr And $^{228}$fr ProduCed iN The SpallatIon of

decreases exponentially w ith the pe riodof21mi nute s $\ pm$ 1.\] Thish alf- life agrees with the p resen tl y ado p te d val ue of 2 2 .0 0 ( 7)mi n. Th is ob serva tio n of $^ {223}$Fr a lso r epresented t h ediscoveryofthe elementfra ncium. $^{ 2 24-22 6}$ Fr {# fr-16. u nnumbe red} ---- -- - ------ - Hanse n et al.  reported the fir s to bservation of$^{224 }$ F r, $ ^{2 25} $Fr and $^ {2 26}$F r in the pa p e r “D e cay character istics of s h ort -lived r adi o -nucli des s tu d ied by on-line iso tope sepa ratort echniqu e s” in 1 969 [@ 196 9Ha n01] . P ro ton so f 6 0 0MeV fro m the CE RN s ynchr ocyc l o t r on b omb arde d a m olten tin tar get and $^{ 224}$ Fr, $ ^{22 5} $Fr a nd $^{ 226}$ Fr were separated usi ng the IS OLD Efac il ity.T he pap ersum marized the IS O LDE p r o g ra m and did not cont ai n de tails ab out th e i nd i vidual n uc lei but t he re sult s w ere pres entedi natable.Th e meas ur edhal f-lif e of2.67(2 0) min f or $^ { 224}$Fr agrees with the pres e nt l y a d opte d v alue of 3.3 3(10 )  min and th e 3 . 9(2)min f or $^ { 225}$Fr is included i n thecalcu lation of the currently a c cepted h alf- l if e of 3.95(14) m in. T he half-li f e of 1.4 3(23)  min for $^{226}$ F r is with ina f act oro f t wo of the pre s e nt v al ue of 4 9(1 ) s. $ ^{2 27, 228 }$F r{#fr-17.u nnumbere d} - -- -- --- ----- - In 197 2Kla pi sch et a l . repo rtedthefi rs t ob servati o no f $^{ 22 7} $Frand $ ^{228 }$Fr in“Half-l ife of th e n e w is ot op e $^{32 }$Na; Observa ti on of $^{3 3} $Na and o t h er new i sotopes produced in the reactio n o f hig h-en ergy prot ons on U” [@ 1 972Kla 01]. U raniu mtar g e ts we r e b omb ar ded with 2 4 GeV prot on s fr om theCERN proton synchr o tro n. $^{227}$Fr an d $^ { 2 28 }$F r w e reid e nti f i ed by on-line m ass spectr om e tr y and deca y cu rv es were measur ed. “ F ollowin g the sam e procedu re asf o r N a, the iso topes $^ {48}$K, $ ^ {49}$ K ,and $ ^{5 0}$K w er e f ound. Howev e r,their half- li ves we re no tshort co mpared with the diffusi on tim e, an d h ence coul d n o t b e determi ned. We also o bse rve d the ne w neut ron- r ic h i s otope s $^ { 227}$Fr a n d$^{ 2 2 8} $Fr produce d i n t he sp all a tion o f

decreases_exponentially with_the period of 21_minutes $\pm$_1.\]_This half-life_agrees_with the presently_adopted value of_22.00(7) min. This observation of_$^{223}$Fr also represented_the_discovery of the element francium. $^{224-226}$Fr {#fr-16.unnumbered} -------------- Hansen et al. reported the first observation of $^{224}$Fr, $^{225}$Fr_and_$^{226}$Fr in_the_paper_“Decay characteristics of short-lived radio-nuclides_studied by on-line isotope separator_techniques” in_1969 [@1969Han01]. Protons of 600 MeV from the CERN_synchrocyclotron_bombarded a molten_tin target and $^{224}$Fr, $^{225}$Fr and $^{226}$Fr were separated_using the ISOLDE facility. The paper_summarized the ISOLDE_program_and_did not contain details_about the individual nuclei but the_results were presented in a table._The measured half-life of 2.67(20) min for $^{224}$Fr_agrees with the presently adopted value_of 3.33(10) min and the 3.9(2) min_for $^{225}$Fr_is included in the calculation_of the currently_accepted half-life_of 3.95(14) min. The_half-life of 1.43(23) min for $^{226}$Fr is_within a factor_of two of the present value_of_49(1) s. $^{227,228}$Fr {#fr-17.unnumbered} -------------- In 1972_Klapisch_et_al. reported the_first observation of_$^{227}$Fr_and $^{228}$Fr_in_“Half-life of the new isotope $^{32}$Na;_Observation_of $^{33}$Na and other new isotopes produced_in the reaction of_high-energy_protons on U” [@1972Kla01]._Uranium targets were bombarded with_24 GeV protons from the CERN_proton synchrotron._$^{227}$Fr and_$^{228}$Fr were identified by on-line mass spectrometry and decay curves were_measured. “Following the same procedure as_for Na, the isotopes_$^{48}$K, $^{49}$K,_and_$^{50}$K were found._However,_their half-lives_were not short compared with the diffusion_time, and_hence could not be determined. We_also observed the new_neutron-rich_isotopes $^{227}$Fr and $^{228}$Fr produced in_the spallation of

}{d\rho}, \\ c_{\Gamma\Theta} &= \frac{\alpha}{3 (3c_{123}+2\alpha \gamma_2)}\left[\left( 1-\frac{2c_{13}}{c_{14}}\right)(1+2\gamma_2) - 6\epsilon_H \gamma_2\left( 1+\frac{2}{3}\gamma_3\right) \right] + \frac{dP}{d\rho}, \\ c_{\Gamma W} &= \frac{2\gamma_1(1+2\gamma_2)}{3\left( 2\gamma_1-1 \right) }, \\ c_{\Gamma X} & = \frac{4\alpha\gamma_1}{3\left( 2\gamma_1-1 \right)(3c_{123}+2\alpha \gamma_2)}\left[\left(1+ \frac{c_{13}}{c_{14}}\right) (1+2\gamma_2)^2 + \frac{3c_{13}}{\alpha}\left(1+2\gamma_2\left[1-\epsilon_H\left(1+\frac{2}{3}\gamma_3 \right) \right] \right) \right], \\ \label{c_GammaY} c_{\Gamma Y} &= \frac{2\gamma_1(1+2\gamma_2)}{9\left( 1-2\gamma_1\right) }.\end{aligned}$$ Note that in and the perturbed fluid variables are those for the dark energy fluid. In order to ensure these results are truly gauge invariant, we must do the same calculation in the synchronous gauge. However, as mentioned previously, we now have an extra variable, $T$, to deal with. Therefore, let us suppose that in the synchronous gauge we find that $$\begin{aligned} w_{\mathrm{de}}\Pi^S &= c_{\Pi \Delta} \Delta + c_{\Pi \Theta}\hat{\Theta} + c_{\Pi X}X + c_{\Pi Y}K^2 Y+c_{\Pi T}T, \\ w_{\mathrm{de}}\Gamma &= c_{\Gamma\Delta} \Delta + c_{\Gamma\Theta}\hat{\Theta} + c_{\Gamma W} W + c_{\Gamma X} X + c_{\Gamma Y} K^

} { d\rho }, \\ c_{\Gamma\Theta } & = \frac{\alpha}{3 (3c_{123}+2\alpha \gamma_2)}\left[\left (1-\frac{2c_{13}}{c_{14}}\right)(1 + 2\gamma_2) - 6\epsilon_H \gamma_2\left (1+\frac{2}{3}\gamma_3\right) \right ] + \frac{dP}{d\rho }, \\ c_{\Gamma W } & = \frac{2\gamma_1(1 + 2\gamma_2)}{3\left (2\gamma_1 - 1 \right) }, \\ c_{\Gamma X } & = \frac{4\alpha\gamma_1}{3\left (2\gamma_1 - 1 \right)(3c_{123}+2\alpha \gamma_2)}\left[\left(1 + \frac{c_{13}}{c_{14}}\right) (1 + 2\gamma_2)^2 + \frac{3c_{13}}{\alpha}\left(1 + 2\gamma_2\left[1-\epsilon_H\left(1+\frac{2}{3}\gamma_3 \right) \right ] \right) \right ], \\ \label{c_GammaY } c_{\Gamma Y } & = \frac{2\gamma_1(1 + 2\gamma_2)}{9\left (1 - 2\gamma_1\right) } .\end{aligned}$$ Note that in and the perturbed fluid variables are those for the blue department of energy fluid. In order to ensure these result are sincerely gauge invariant, we must do the like calculation in the synchronous gauge. However, as mentioned previously, we now suffer an extra variable, $ T$, to distribute with. Therefore, let us speculate that in the synchronous bore we find that $ $ \begin{aligned } w_{\mathrm{de}}\Pi^S & = c_{\Pi \Delta } \Delta + c_{\Pi \Theta}\hat{\Theta } + c_{\Pi X}X + c_{\Pi Y}K^2 Y+c_{\Pi T}T, \\ w_{\mathrm{de}}\Gamma & = c_{\Gamma\Delta } \Delta + c_{\Gamma\Theta}\hat{\Theta } + c_{\Gamma W } W + c_{\Gamma X } X + c_{\Gamma Y } K^

}{d\rhl}, \\ c_{\Gamma\Theta} &= \frac{\alpha}{3 (3c_{123}+2\alpha \gamma_2)}\leyr[\left( 1-\hrac{2c_{13}}{c_{14}}\rjght)(1+2\gammx_2) - 6\epsilon_H \gamma_2\left( 1+\frac{2}{3}\gemma_3\eight) \right] + \frac{dP}{d\rho}, \\ c_{\Eamma W} &= \vrac{2\gammq_1(1+2\gamnq_2)}{3\left( 2\gamme_1-1 \right) }, \\ c_{\Gammz X} & = \hrac{4\alpha\gamma_1}{3\lgft( 2\gamma_1-1 \richt)(3c_{123}+2\alpha \gammd_2)}\ldfc[\left(1+ \frac{c_{13}}{c_{14}}\right) (1+2\gamma_2)^2 + \frac{3c_{13}}{\alpha}\lqft(1+2\gamms_2\lfft[1-\epsilon_H\lefj(1+\frac{2}{3}\bwmma_3 \ginht) \right] \right) \right], \\ \label{c_GajmaY} c_{\Gakma Y} &= \frac{2\gakma_1(1+2\gamma_2)}{9\left( 1-2\gamma_1\right) }.\enf{alihned}$$ Note that in wnd the perjhrbqe fluid varixbles are uhmse for thg dark energy fluid. In order to evsure these resolfs dre truly geuge igvariant, we must do dhe samr calculation ln thx sybchronous gauge. Howevxr, as mentioned prevyously, we nkw have an extra caeiablg, $T$, tm dexo wkth. Tiersfore, pet us supposs that in tye synchronous gaugt wq find that $$\befin{alidnqd} w_{\mathrm{de}}\Pi^S &= c_{\Pi \Delta} \Delta + c_{\Pi \Thtta}\haf{\Theta} + c_{\Pi X}X + c_{\Pi Y}K^2 T+c_{\Pi T}T, \\ w_{\mathrm{de}}\Gammw &= c_{\Gamma\Qelta} \Delta + c_{\Gamma\Theta}\hat{\Theta} + c_{\Gamma W} W + c_{\Gakma X} B + e_{\Nwona Y} K^

}{d\rho}, \\ c_{\Gamma\Theta} &= \frac{\alpha}{3 (3c_{123}+2\alpha \gamma_2)}\left[\left( 6\epsilon_H 1+\frac{2}{3}\gamma_3\right) \right] \frac{dP}{d\rho}, \\ c_{\Gamma }, c_{\Gamma X} & \frac{4\alpha\gamma_1}{3\left( 2\gamma_1-1 \right)(3c_{123}+2\alpha \frac{c_{13}}{c_{14}}\right) (1+2\gamma_2)^2 + \frac{3c_{13}}{\alpha}\left(1+2\gamma_2\left[1-\epsilon_H\left(1+\frac{2}{3}\gamma_3 \right) \right] \right], \\ \label{c_GammaY} c_{\Gamma Y} &= \frac{2\gamma_1(1+2\gamma_2)}{9\left( 1-2\gamma_1\right) }.\end{aligned}$$ Note that in and perturbed fluid variables are those for the dark energy fluid. In order to these are gauge we must do the same calculation in the synchronous gauge. However, as mentioned previously, we now an extra variable, $T$, to deal with. Therefore, us suppose that in synchronous gauge we find that w_{\mathrm{de}}\Pi^S c_{\Pi \Delta} + \Theta}\hat{\Theta} c_{\Pi X}X + Y}K^2 Y+c_{\Pi T}T, \\ w_{\mathrm{de}}\Gamma &= c_{\Gamma\Delta} \Delta + c_{\Gamma\Theta}\hat{\Theta} + c_{\Gamma W} W + c_{\Gamma X} + c_{\Gamma

}{d\rho}, \\ c_{\Gamma\Theta} &= \frac{\alpha}{3 (3c_{123}+2\Alpha \gamma_2)}\Left[\lEft( 1-\FraC{2c_{13}}{C_{14}}\rigHt)(1+2\gaMma_2) - 6\epsilon_H \gamMA_2\lefT( 1+\frac{2}{3}\gamma_3\right) \right] + \frAc{dP}{d\RhO}, \\ C_{\GamMA W} &= \Frac{2\gAmma_1(1+2\gamMA_2)}{3\lEFT( 2\gaMmA_1-1 \rIghT) }, \\ c_{\gAmMa X} & = \frAc{4\aLpha\gamMa_1}{3\left( 2\gammA_1-1 \riGhT)(3c_{123}+2\alpha \gamma_2)}\LEfT[\left(1+ \frac{c_{13}}{C_{14}}\riGht) (1+2\gamma_2)^2 + \frac{3C_{13}}{\alPha}\lefT(1+2\gAmmA_2\Left[1-\ePsiLon_H\lEft(1+\fraC{2}{3}\Gamma_3 \rIght) \right] \RiGHt) \righT], \\ \Label{c_GAMMay} c_{\GaMma Y} &= \frac{2\gamma_1(1+2\gammA_2)}{9\LeFT( 1-2\gamma_1\right) }.\end{AligneD}$$ NOTe THAt iN anD the perturBeD fluiD VariablES aRE THosE For the dark eneRgy fluid. In oRDer To ensuRe TheSE resulTs are TrULy gAuge invariaNt, we Must do the Same caLCulatioN In the syNchronOus GauGe. HoWEvEr, As mEnTIonED pRevIOusLy, we now hAvE aN extrA varIABLE, $T$, to DeaL witH. TherEfore, let us supPosE thaT In tHe synChronOus gAuGe we fInd thaT $$\begiN{aLigned} w_{\mathrm{de}}\pi^S &= c_{\pi \Delta} \DeLta + C_{\PI \ThEtA}\hat{\THEta} + c_{\Pi x}X + c_{\pi Y}k^2 Y+c_{\Pi T}T, \\ W_{\mathrm{DE}}\GaMmA &= C_{\gAmMa\Delta} \Delta + c_{\Gamma\thETA}\hAt{\Theta} + c_{\gamma W} w + C_{\GAmMA X} X + c_{\GammA Y} k^

}{d\rho}, \\ c_{\Gamma\The ta} &= \fr ac{\a lph a}{ 3(3c_ {123 }+2\alpha \gam m a_2) }\left[\left( 1-\frac{ 2c_{1 3} } {c_{ 1 4} }\rig ht)(1+2 \ ga m m a_2 ) - 6\ ep s il on_H\ga mma_2\l eft( 1+\fr ac{ 2} {3}\gamma_3\ r ig ht) \rig ht] + \frac{dP} {d\ rho},\\ c_ { \Gamm a W } &=\frac{ 2 \gamma _1(1+2\ga mm a _2)}{3 \ left( 2 \ g am ma_1 -1 \right) }, \\c _{ \ Gamma X} & = \ frac{4 \a l ph a \ gam ma_ 1}{3\left( 2 \gamm a _1-1 \r i gh t ) ( 3c_ { 123}+2\alpha\gamma_2)}\ l eft [\left (1 + \ f rac{c_ {13}} {c _ {14 }}\right) ( 1+2\ gamma_2)^ 2 + \f r ac{3c_{ 1 3}}{\al pha}\l eft (1+ 2\ga m ma _2 \le ft [ 1-\ e ps ilo n _H\ left(1+\ fr ac {2}{3 }\ga m m a _ 3 \r igh t) \righ t] \right) \r igh t],\ \ \ label {c_Ga mmaY }c_{\G amma Y } &=\f rac{2\gamma_1(1 +2\g amma_2)}{ 9\l ef t(1- 2\gam m a_1\ri ght ) } .\end{a ligned} $ $ N ot e t ha t in and the pertu rb e d f luid var iables ar et hose for t hedark e nergy flu i d. In ord er toe ns ur e these r esults a retru ly ga u ge i nvaria nt, we m ust d o the same calc u lation in the sy n c hr o nous ga uge. Howeve r, a s men tion e dpre v iousl y, we n o wh ave an extra variab le , $T$, to d eal with. The refore, le t u s suppos e th a ti n the synchron ous g auge we fi n d that $ $\beg in{align ed} w_{\m a t hrm{de}} \Pi ^S&=c_{ \ P i\Delta} \Delt a + c_ {\ Pi \The ta} \hat{\T het a}+ c _{\ Pi X}X + c_ {\Pi Y}K ^2 Y +c _{ \Pi T}T, \\ w_{\m at hrm {d e}} \Gamm a &= c_ {\Gam ma\D el ta } \D elta +c _{ \ G amma \T he ta}\ hat {\ Theta } +c _{\ Gamma W } W + c_{ \Ga m ma X }X+ c_{\G amma Y} K^

}{d\rho}, \\ c_{\Gamma\Theta}_&= \frac{\alpha}{3_(3c_{123}+2\alpha \gamma_2)}\left[\left( 1-\frac{2c_{13}}{c_{14}}\right)(1+2\gamma_2) _- 6\epsilon_H_\gamma_2\left(_1+\frac{2}{3}\gamma_3\right) __\right] + \frac{dP}{d\rho},_\\ c_{\Gamma W} &=_\frac{2\gamma_1(1+2\gamma_2)}{3\left( 2\gamma_1-1 \right) },_\\ c_{\Gamma X} &_=_\frac{4\alpha\gamma_1}{3\left( 2\gamma_1-1 \right)(3c_{123}+2\alpha \gamma_2)}\left[\left(1+ \frac{c_{13}}{c_{14}}\right) (1+2\gamma_2)^2 + \frac{3c_{13}}{\alpha}\left(1+2\gamma_2\left[1-\epsilon_H\left(1+\frac{2}{3}\gamma_3 \right) \right] \right) \right], \\_\label{c_GammaY} c_{\Gamma_Y} &=_\frac{2\gamma_1(1+2\gamma_2)}{9\left(_1-2\gamma_1\right)_}.\end{aligned}$$ Note that in and_the perturbed fluid variables are_those for_the dark energy fluid. In order to ensure these_results_are truly gauge_invariant, we must do the same calculation in the_synchronous gauge. However, as mentioned previously,_we now have_an_extra_variable, $T$, to deal_with. Therefore, let us suppose that_in the synchronous gauge we find_that $$\begin{aligned} w_{\mathrm{de}}\Pi^S &= c_{\Pi \Delta} \Delta +_c_{\Pi \Theta}\hat{\Theta} + c_{\Pi X}X +_c_{\Pi Y}K^2 Y+c_{\Pi T}T, \\ w_{\mathrm{de}}\Gamma_&= c_{\Gamma\Delta}_\Delta + c_{\Gamma\Theta}\hat{\Theta} + c_{\Gamma_W} W +_c_{\Gamma X}_X + c_{\Gamma_Y} K^

otimes E) \preceq \phi(S' \otimes E)$. \[def:adaptation-opt\] Given two systems $S$ and $S'$, $S'$ is at least as adaptive as $S$ with respect to optimization, written $S \sqsubseteq^* S'$ iff for all adaptation domains $\mathcal{A} \in \mathfrak{A}$ it holds that $S \Vdash \mathcal{A} \Longrightarrow S' \Vdash \mathcal{A}$ and $S \preceq_\mathcal{A} S'$. Note that so far our notions of adaptivity and optimization are purely extensional, which originates from the black box perspective on adaptation assumed in [@holzl2011towards]. Constructing Adaptive Systems ----------------------------- We now shift the focus of our analysis a bit away from the question “When is a system adaptive?” towards the question “How is a system adaptive?”. This refers to both questions of software architecture (i.e., which components should we use to make an adaptive system?) and questions of software engineering (i.e., which development processes should we use to develop an adaptive system?). We will see that with the increasing usage of methods of artificial intelligence, design-time engineering and run-time adaptation increasingly overlap [@wirsing2015software]. \[def:adaptation-sequence\] A series of $|I|$ systems $\mathcal{S} = (S_i)_{i\in I}$ with index set $I$ with a preorder $\leq$ on the elements of $I$ is called an *adaptation sequence* iff for all $i, j \in I$ it holds that $i \leq j \Longrightarrow S_i \sqsubseteq^* S_j$ Note that we used adaptation with optimization in Definition \[def:adaptation-sequence\] so that a sequence of systems $(S_i)_{i\in I}$ that each fulfill the same hard constraints ($\gamma$ within a singleton adaptation space $\mathfrak{A} = \{\{(E, \gamma, \phi)\}\}$) can form an adaptation sequence iff for all $i, j \in I$ it holds that $i \leq j \Longrightarrow \phi(S_i \otimes E) \preceq \phi(S_j \otimes E)$. This is the purest formulation of an optimization process within our formal framework.[^2] Such an adaptation sequence can

otimes E) \preceq \phi(S' \otimes E)$. \[def: adaptation - opt\ ] Given two systems $ S$ and $ S'$, $ S'$ is at least as adaptive as $ S$ with respect to optimization, publish $ south \sqsubseteq^ * S'$ iff for all adaptation domains $ \mathcal{A } \in \mathfrak{A}$ it holds that $ S \Vdash \mathcal{A } \Longrightarrow S' \Vdash \mathcal{A}$ and $ S \preceq_\mathcal{A } S'$. notice that so far our notions of adaptivity and optimization are purely extensional, which originate from the black box perspective on adaptation assumed in [ @holzl2011towards ]. Constructing Adaptive Systems ----------------------------- We nowadays shift the focus of our psychoanalysis a bit away from the question “ When is a system adaptive? ” towards the doubt “ How is a system adaptive? ”. This refers to both questions of software architecture (i.e., which components should we practice to make an adaptive system ?) and questions of software technology (i.e., which development processes should we use to develop an adaptive system ?). We will understand that with the increasing usage of methods of artificial intelligence, design - meter engineering and run - time adaptation increasingly overlap [ @wirsing2015software ]. \[def: adaptation - sequence\ ] A series of $ |I|$ systems $ \mathcal{S } = (S_i)_{i\in I}$ with index set $ I$ with a preorder $ \leq$ on the elements of $ I$ is called an * adaptation sequence * iff for all $ i, j \in I$ it holds that $ i \leq j \Longrightarrow S_i \sqsubseteq^ * S_j$ Note that we used adaptation with optimization in Definition   \[def: adaptation - sequence\ ] so that a sequence of systems $ (S_i)_{i\in I}$ that each fulfill the same hard constraint ($ \gamma$ within a singleton adaptation space $ \mathfrak{A } = \{\{(E, \gamma, \phi)\}\}$) can form an adaptation succession iff for all $ i, j \in I$ it hold that $ i \leq j \Longrightarrow \phi(S_i \otimes E) \preceq \phi(S_j \otimes E)$. This is the saturated formulation of an optimization process within our formal framework.[^2 ] Such an adaptation sequence can

otiles E) \preceq \phi(S' \otimes E)$. \[def:adaptation-opt\] Givxn two aystems $R$ and $S'$, $S'$ is at least as ada'tivw as $W$ with respect to optioization, aritten $W \sqwybseteq^* S'$ mrf for all adzitatimi domains $\mathcsl{A} \in \matvfrak{A}$ it holdv ghct $S \Vdash \mathcal{A} \Longrightarrow S' \Vdash \kahhcal{A}$ and $S \pteceq_\kwthczl{A} S'$. Note that so far our notions kf adapuivity and optimizstion are purely extensionwl, wjich originates frlm the blacj bov perspective on adaptauimn assumed in [@holzl2011towards]. Constructing Adxptivz Systems ----------------------------- We niw dvift the forus of our analysis a bit dway frpm the questiok “Whei is a system adaptive?” toxards the question “Hjw is a sfscem adaptive?”. This refwrw to toth querriovs kf skftwarf acchitecture (i.e., which cimponents should we ufv to make an zdaptide system?) and questions of software enginterinf (i.e., which development processes should we ude to devqlop an adaptive system?). We will see that with the incrxariny usagd ov methods of artificial intelligence, design-tiis tngpneering and run-tlme adaptation inctewsoggly overlap [@dirsiny2015aortware]. \[def:adaptatioj-sequense\] A weries of $|I|$ sustems $\mathcal{S} = (S_i)_{i\in I}$ wirh index set $U$ with a preorder $\peq$ on the zlemenjs of $O$ is called an *adaptatiun ssquence* iff for all $j, j \in I$ it holds thst $i \leq j \Longrightarrow S_i \sqsubsetxq^* S_j$ Uote thag we used wdaptation with optimization in Devinitnon \[dex:adaptatioj-sequence\] so that a sequence of systems $(S_i)_{i\in I}$ thdt each yulfilk the same hawd constraints ($\gamma$ cithin a singletkn adapvation space $\mathfrak{A} = \{\{(E, \hamma, \phi)\}\}$) cai form an adaptatuon seqjdnce iff for akl $i, j \in I$ it holdw that $i \leq j \Lonnrighjadrow \phi(S_i \otimzr E) \preceq \phi(S_j \otkmef V)$. Tiis if the purest xormjlagoon ow an opunmidatkon lrocess within our fmrmam framework.[^2] Such an aqaptatiob sequense can

otimes E) \preceq \phi(S' \otimes E)$. \[def:adaptation-opt\] systems and $S'$, is at least respect optimization, written $S S'$ iff for adaptation domains $\mathcal{A} \in \mathfrak{A}$ it that $S \Vdash \mathcal{A} \Longrightarrow S' \Vdash \mathcal{A}$ and $S \preceq_\mathcal{A} S'$. Note so far our notions of adaptivity and optimization are purely extensional, which originates the box on assumed in [@holzl2011towards]. Constructing Adaptive Systems ----------------------------- We now shift the focus of our analysis a away from the question “When is a system towards the question “How a system adaptive?”. This refers both of software (i.e., components we use to an adaptive system?) and questions of software engineering (i.e., which development processes should we use to develop adaptive system?). see that the usage methods of artificial engineering and run-time adaptation increasingly overlap series of $|I|$ systems $\mathcal{S} = (S_i)_{i\in I}$ index set with a preorder $\leq$ on the of $I$ is called an *adaptation sequence* iff all $i, j \in I$ it holds that $i \leq j \Longrightarrow S_i \sqsubseteq^* S_j$ we used adaptation with in Definition \[def:adaptation-sequence\] that sequence systems I}$ that fulfill the same hard constraints ($\gamma$ within a singleton adaptation space = \{\{(E, \gamma, \phi)\}\}$) can form an adaptation sequence iff $i, \in I$ it that $i \leq j \phi(S_i E) \preceq \phi(S_j \otimes is purest optimization within formal framework.[^2] Such an sequence can

otimes E) \preceq \phi(S' \otimes E)$. \[dEf:adaptatiOn-opt\] givEn tWo SystEms $S$ And $S'$, $S'$ is at least AS adaPtive as $S$ with respect to oPtimiZaTIon, wRItTen $S \sQsubsetEQ^* S'$ IFF foR aLl AdaPtATiOn domAinS $\mathcaL{A} \in \mathfrAk{A}$ It Holds that $S \VdASh \Mathcal{A} \LoNgrIghtarrow S' \VdAsh \MathcaL{A}$ And $s \PreceQ_\maThcal{a} S'$. Note THat so fAr our notiOnS Of adapTIvity anD OPtImizAtion are purely extENsIOnal, which origiNates fRoM ThE BLacK boX perspectiVe On adaPTation aSSuMED In [@hOLzl2011towards]. ConStructing AdAPtiVe SystEmS ----------------------------- We NOw shifT the fOcUS of Our analysis A bit Away from tHe quesTIon “When IS a systeM adaptIve?” TowArds THe QuEstIoN “how IS a SysTEm aDaptive?”. THiS rEfers To boTH QUEstiOns Of soFtwarE architecture (I.e., wHich COmpOnentS shouLd we UsE to maKe an adAptivE sYstem?) and questioNs of Software eNgiNeEriNg (I.e., whiCH develOpmEnt ProcessEs shoulD We uSe TO DEvElop an adaptive systEm?). wE WiLl see thaT with tHE iNcREasing usAgE of MethODS of arTifiCIaL intelliGence, dESiGn-Time engInEering AnD ruN-tiMe adaPTatiOn incrEasingly OverlAP [@wirsing2015softwaRE]. \[def:adaptatioN-SeQUEnCE\] A seRieS of $|I|$ systems $\MathCAl{S} = (S_I)_{i\in i}$ WiTh iNDex seT $I$ witH a PReORder $\leq$ on the elementS oF $I$ is caLled aN *adaptation seQuence* iff fOR ALl $i, j \in I$ iT holDS tHAt $i \leq j \LongrigHtarrOw S_i \sqsubsETeq^* S_j$ NotE that We used adAptation wITH optimizAtiOn iN DeFinITIoN \[def:adaptatioN-SEqueNcE\] so that A seQuence oF sySteMs $(S_I)_{i\iN I}$ That each fUlfill thE sAmE hArD coNstraINts ($\gamma$ WiThiN a SinGletoN AdaptaTion sPace $\MaThFRak{a} = \{\{(E, \gamma, \PHi)\}\}$) CAN forM aN aDaptAtiOn SequeNce iFF foR all $i, j \iN I$ it holds ThaT $I \leq J \LOnGrightaRrow \phi(S_i \otimEs e) \preceq \phi(s_j \OtiMes E)$. ThIS Is the purEst formulation of an optimIZation pRocEss wiThin Our formal FraMework.[^2] sucH An adapTation SequeNcE caN

otimes E) \preceq \phi(S'\otimes E) $. \ [de f:a da ptat ion- opt\] Given tw o sys tems $S$ and $S'$, $S' $ isat leas t a s ada ptive a s $ S $ wi th r esp ec t t o opt imi zation, written $ S \ sq subseteq^* S ' $iff for al l a daptation do mai ns $\m at hca l {A} \ in\math frak{A } $ it h olds that $ S \Vdas h \mathc a l {A } \L ongrightarrow S'\ Vd a sh \mathcal{A} $ and$S \p r e ceq _\m athcal{A}S' $. N o te that so f a r o u r notions ofadaptivitya ndoptimi za tio n are p urely e x ten sional, whi ch o riginates fromt he blac k box pe rspect ive on ada p ta ti onas s ume d i n [ @ hol zl2011to wa rd s]. Cons t r u c ting Ad apti ve Sy stems ------- --- ---- - --- ----- ----- - W enow s hift t he fo cu s of our analys is a bit away fr om th equest i on “Wh enisa syste m adapt i ve? ”t o w ar ds the question “H ow i sa system adapt i ve ?” . This re fe rsto b o t h que stio n sof softw are ar c hi te cture ( i. e., wh ic h c omp onent s sho uld we use tomakea n adaptive sys t em?) and ques t io n s o f sof twa re engineer ing( i.e. , wh i ch de v elopm ent p ro c es s es should we use to d evelop an a daptive syste m?). We wi l l see that wit h t h e increasing u sageof methods of artif icial intelli gence, de s i gn-timeeng ine eri nga n drun-time adap t a tion i ncreasi ngl y overl ap[@w irs ing 20 15softwar e]. \[d ef :a da pt ati on-se q uence\]Aser ie s o f $|I | $ syst ems $ \mat hc al { S}= (S_i) _ {i \ i n I} $wi th i nde xset $ I$ w i tha preor der $\leq $ o n the e le ments o f $I$ is call ed an *adapt at ion seque n c e* iff f or all $i, j \in I$ ith olds th at$i \l eq j \Longrig hta rrow S _i\ sqsubs eteq^* S_j$ Not e thatw e u sed a daptationw i thoptim iz atio n in De finition \[def:ada p tat ion-sequence\ ] s o th a t a se q ue n ceof sys t e ms $(S_i)_{i\in I}$ thatea c hfulfill th e sa me hard c onstrai nts ( $ \gamma$ within a singleto nadap t a tio n space $\ mathfrak {A} = \{\ { (E, \ g am ma, \ phi )\}\}$ )can form an ad a pta tionsequen ce iff f or al l$i, j \i n I$ it holds that $i \ leq j\Long rig htarrow \ phi ( S_i \otimesE) \ preceq \ph i(S _j\otim esE )$. T hisi sthe pures t fo r mulationo fano p ti mization pr o c e sswithi n o u r form al f ramework.[^2] Su c h an adaptatio n se q u enc e c a n

otimes E)_\preceq \phi(S'_\otimes E)$. \[def:adaptation-opt\] Given two_systems $S$_and_$S'$, $S'$_is_at least as_adaptive as $S$_with respect to optimization,_written $S \sqsubseteq^*_S'$_iff for all adaptation domains $\mathcal{A} \in \mathfrak{A}$ it holds that $S \Vdash \mathcal{A}_\Longrightarrow_S' \Vdash_\mathcal{A}$_and_$S \preceq_\mathcal{A} S'$. Note that so_far our notions of adaptivity_and optimization_are purely extensional, which originates from the black_box_perspective on adaptation_assumed in [@holzl2011towards]. Constructing Adaptive Systems ----------------------------- We now shift the focus_of our analysis a bit away_from the question_“When_is_a system adaptive?” towards_the question “How is a system_adaptive?”. This refers to both questions_of software architecture (i.e., which components should_we use to make an adaptive_system?) and questions of software_engineering (i.e.,_which development processes should we_use to develop_an adaptive_system?). We will_see that with the increasing usage_of methods of_artificial intelligence, design-time engineering and run-time_adaptation_increasingly overlap [@wirsing2015software]. \[def:adaptation-sequence\]_A_series_of $|I|$_systems $\mathcal{S} =_(S_i)_{i\in_I}$ with_index_set $I$ with a preorder $\leq$_on_the elements of $I$ is called an_*adaptation sequence* iff for_all_$i, j \in I$_it holds that $i \leq_j \Longrightarrow S_i \sqsubseteq^* S_j$ Note that_we used_adaptation with_optimization in Definition \[def:adaptation-sequence\] so that a sequence of systems $(S_i)_{i\in I}$_that each fulfill the same hard_constraints ($\gamma$ within a_singleton adaptation_space_$\mathfrak{A} = \{\{(E,_\gamma,_\phi)\}\}$) can_form an adaptation sequence iff for all_$i, j_\in I$ it holds that $i_\leq j \Longrightarrow \phi(S_i_\otimes_E) \preceq \phi(S_j \otimes E)$. This_is the purest formulation of an_optimization process within our formal_framework.[^2] Such_an_adaptation sequence can

of blue HB and RR Lyrae variable stars in And I is that they represent a (small) population of stars that formed perhaps 3 Gyr or more before the bulk of the stellar population in And I. Such a result implies that, just as is the case for many of the galactic dSphs, star formation in And I occurred over an extended interval. The driving mechanism(s) behind the extended and diverse star formation histories of the Local Group dSph systems remains one of the most outstanding unsolved problems in this branch of Astronomy. Horizontal Branch Morphology Radial Gradients --------------------------------------------- A further property of And I that is worthy of additional comment is that established in section 3.1.2: the existence of a radial gradient in the morphology of And I’s horizontal branch. The sense of the gradient is such that there are relatively more blue horizontal branch stars outside the core radius. This gradient could reflect either an underlying abundance gradient or a radial age gradient, but given the above discussion, it would seem more likely that it should be interpreted as suggesting that And I was more extended at its initial star forming epoch, than when the bulk of its stars formed. In order to shed more light on this possibility, we have sought the existence of similar gradients in other dSph galaxies. This task requires dSph studies that fulfill the following criteria: a) the available photometry must extend beyond the core radius; b) the photometry must be reasonably precise at the magnitude of the horizontal branch; and c) the HB morphology must contain both blue and red stars. These criteria are satisfied by three studies: that of Carina by Smecker-Hane [*et al. *]{}(1994), of Leo II by Demers & Irwin (1993), and of Sculptor by Light (1988). Considering first Carina, the extensive study of Smecker-Hane [*et al. *]{}(1994) clearly separates the “red clump” of intermediate-age core helium burning stars from the horizontal branch of old stars (both red and blue). Smecker-Hane [*et al. *]{}(1994) find no difference in the surface density distributions of these two classes of stars and further, they indicate that there is no systematic trend with radius in the fraction of the total core helium burning star population represented by the old HB stars. These results apply out to a radius of $\sim$10$\arcmin$ beyond which their sampling is incomplete. Using

of blue HB and RR Lyrae variable stars in And I is that they represent a (humble) population of star that formed perhaps 3 Gyr or more before the bulk of the leading population in And I. Such a result implies that, just as is the font for many of the galactic dSphs, star constitution in And I occurred over an extended time interval. The driving mechanism(s) behind the extended and diverse star constitution histories of the Local Group dSph systems remains one of the most outstanding unsolved problem in this branch of Astronomy. Horizontal Branch Morphology Radial Gradients --------------------------------------------- A further property of And I that is worthy of extra comment is that established in part 3.1.2: the existence of a radial gradient in the morphology of And I ’s horizontal outgrowth. The sense of the gradient is such that there are relatively more blue horizontal branch stars outside the effect radius. This gradient could reflect either an underlying abundance gradient or a radial age gradient, but given the above discussion, it would seem more likely that it should be interpreted as suggesting that And I was more extended at its initial star forming epoch, than when the bulk of its stars formed. In order to shed more luminosity on this possibility, we have sought the universe of exchangeable gradients in other dSph galaxies. This undertaking requires dSph studies that fulfill the following criteria: a) the available photometry must prolong beyond the core radius; b) the photometry must be reasonably precise at the magnitude of the horizontal branch; and c) the HB morphology must hold both blue and red stars. These criteria are quenched by three studies: that of Carina by Smecker - Hane [ * et al.   * ] { } (1994), of Leo II by Demers & Irwin (1993), and of Sculptor by Light (1988). consider inaugural Carina, the extensive study of Smecker - Hane [ * et al.   * ] { } (1994) clearly separate the “ red clump ” of intermediate - old age core helium burning stars from the horizontal branch of honest-to-god stars (both red and gloomy). Smecker - Hane [ * et al.   * ] { } (1994) find no deviation in the surface concentration distributions of these two classes of star and further, they indicate that there be no systematic vogue with radius in the fraction of the total core helium burning star population represented by the old HB stars. These results use out to a radius of $ \sim$10$\arcmin$ beyond which their sampling is incomplete. use

of blue HB and RR Lyrae vaviable stars in Cbd I iv that they reoresent a (small) population oh stqrs tyat formed perhaps 3 Gyf or more before rhe uulk of the stellar population ln Anb M. Such a result implies tvat, just as is tfe case for many of the galactic dSphf, star gogmation in And I obctrres over an extended interval. The drjving mtchanism(s) behind tne extended and diverse stwr flrmation histories of the Locql Gwiup dSph sysgems remains one of thg most outstanding unsolved probuems nn this brabcy ov Astronomy. Hirizogtal Branch Morpholocy Radisl Gradients --------------------------------------------- A nurthxr peoperty of And I that is worthy of addijional comkeut is that establishee un sewtiot 3.1.2: tfw ebisuenre kf a rwdiel gradient in the morphology of And I’s hprydpntal branch. The sqnfe of the gradient is such that there age rslatively more blue horuzontal branch stars lutside tre core radius. This gradient could reflect either an uiddrlvlkg acynfance gradient or a radial age gradient, but gybem nhe above discusslon, it would seem koge jikely that ij shoulb bs interpreted as skggestigg thqt And I ras kore extended at its initiao star formiug wpoch, than when thz bulk of itr stsrs fprmed. In order to shed mure might on thls possibjuity, we have soueht tve existtvce of similar grwdients ii othzr dSph ealacies. Tris task rfquircv dSph studies thah fulyill dhe followlng criteria: a) the available photometry must gxtand beyond che cove radius; b) the photometry muxt be rzasonacly precisv at the kagnitude os the horizondwl branch; anv c) the HF moephooogy murg contain both blue and red stars. These criteria arc satkafied by three wtueies: that of Catinx br Dmxcker-Rdne [*et al. *]{}(1994), of Leo II ny Deoers & Irwin (1993), and of Sculptor by Lighd (1988). Cohsidering first Catika, the exjensive seudy of Smeckrr-Hane [*et al. *]{}(1994) clearpy se'arater the “req clump” of intermediate-age cors helium hurking stars frjm tme hjrizontal yranch of old stars (both red and blue). Smxcker-Hane [*et al. *]{}(1994) find no difference in the xmrface densivy diseributionv of these two classws of stars and fmrther, they indicate tgat thare id no systematic trend with radius in the fraction of the total core heliun burnmnd star popumatipn reirefznted br thx uld HB stars. Thexe results apply out to a radiur of $\sim$10$\dremin$ beyond which their samplone is incompleje. Using

of blue HB and RR Lyrae variable And is that represent a (small) perhaps Gyr or more the bulk of stellar population in And I. Such result implies that, just as is the case for many of the galactic star formation in And I occurred over an extended interval. The driving mechanism(s) the and star histories of the Local Group dSph systems remains one of the most outstanding unsolved problems in branch of Astronomy. Horizontal Branch Morphology Radial Gradients A further property of I that is worthy of comment that established section the of a radial in the morphology of And I’s horizontal branch. The sense of the gradient is such that there relatively more branch stars the radius. gradient could reflect underlying abundance gradient or a radial given the above discussion, it would seem more that it be interpreted as suggesting that And was more extended at its initial star forming than when the bulk of its stars formed. In order to shed more light on we have sought the of similar gradients other galaxies. task dSph studies fulfill the following criteria: a) the available photometry must extend beyond core radius; b) the photometry must be reasonably precise at of horizontal branch; and the HB morphology must both and red stars. These satisfied three Carina Smecker-Hane al. *]{}(1994), of Leo by Demers & Irwin (1993), of Sculptor by Light extensive study of Smecker-Hane [*et al. *]{}(1994) clearly the “red clump” of intermediate-age core helium stars from the horizontal branch of old stars (both red and blue). [*et al. no difference in the surface density distributions of two classes of stars further, they indicate that there is no systematic trend radius the fraction the total core burning star population by the old These results out a beyond which their sampling is incomplete.

of blue HB and RR Lyrae variablE stars in AnD I is tHat TheY rEpreSent A (small) populatiON of sTars that formed perhaps 3 GYr or mOrE BefoRE tHe bulK of the sTElLAR poPuLaTioN iN anD I. SucH a rEsult imPlies that, jUst As Is the case for MAnY of the galaCtiC dSphs, star foRmaTion in anD I oCCurreD ovEr an eXtendeD IntervAl. The drivInG MechanISm(s) behiND ThE extEnded and diverse stAR fORmation historiEs of thE LOCaL gRouP dSPh systems rEmAins oNE of the mOSt OUTStaNDing unsolved pRoblems in thIS brAnch of asTroNOmy. HorIzontAl bRanCh MorphologY RadIal GradieNts --------------------------------------------- A fuRTher proPErty of ANd I thaT is WorThy oF AdDiTioNaL ComMEnT is THat EstablisHeD iN sectIon 3.1.2: tHE EXIsteNce Of a rAdial Gradient in the MorPholOGy oF And I’S horiZontAl BrancH. The seNse of ThE gradient is such That There are rElaTiVelY mOre blUE horizOntAl bRanch stArs outsIDe tHe CORE rAdius. This gradient cOuLD ReFlect eitHer an uNDeRlYIng abundAnCe gRadiENT or a rAdiaL AgE gradienT, but giVEn ThE above dIsCussioN, iT woUld Seem mORe liKely thAt it shouLd be iNTerpreted as sugGEsting that And i WaS MOrE ExteNdeD at its initiAl stAR forMing EPoCh, tHAn wheN the bUlK Of ITs stars formed. In ordeR tO shed mOre liGht on this possIbility, we hAVE Sought thE exiSTeNCe of similar graDientS in other dSPH galaxieS. This Task requIres dSph sTUDies that FulFilL thE foLLOwIng criteria: a) tHE AvaiLaBle photOmeTry must ExtEnd BeyOnd ThE core radiUs; b) the phOtOmEtRy MusT be reASonably pReCisE aT thE magnITude of The hoRizoNtAl BRanCh; and c) tHE Hb MOrphOlOgY musT coNtAin boTh blUE anD red staRs. These crIteRIa arE sAtIsfied bY three studies: ThAt of Carina By smeCker-HaNE [*Et al. *]{}(1994), of LeO II by Demers & Irwin (1993), and of ScULptor by ligHt (1988). ConSideRing first carIna, the ExtENsive sTudy of smeckEr-hanE [*ET al. *]{}(1994) clEARlY sePaRates the “reD CLumP” of inTeRmedIate-age Core helium burning sTArs From the horizoNtaL braNCH oF olD StARs (bOtH Red AND blue). Smecker-HanE [*et al. *]{}(1994) find nO dIFfErence in thE SurFaCe densiTy distrIbutiONs of theSe two clasSes of starS aNd fuRTHer, They indicaTe that thEre is no sySTematIC tRend wIth Radius In The FractIon of tHE toTal coRe heliUm BurninG star PoPulation Represented by the old HB stArs. TheSe resUltS apply out To a RAdiUs of $\sim$10$\arCmin$ Beyond whicH thEir SamplIng IS incoMpleTE. USinG

of blue HB and RR Lyrae v ariable st ars i n A ndIis t hatthey represent a (s mall) population of st ars t ha t for m ed perh aps 3 G y ro r mo re b efo re th e bul k o f the s tellar pop ula ti on in And I. Su ch a resul t i mplies that, ju st asis th e case fo r man y of t h e gala ctic dSph s, star f o rmation i nAndI occurred over a n e x tended interva l. The d r iv i n g m ech anism(s) b eh ind t h e exten d ed a n d d i verse star fo rmation his t ori es ofth e L o cal Gr oup d Sp h sy stems remai ns o ne of the mosto utstand i ng unso lved p rob lem s in th is br an c h o f A str o nom y. Hori zo nt al Br anch M o r phol ogy Rad ial G radients ---- --- ---- - --- ----- ----- ---- -- ----- ------ --- Afurther propert y of And I th atis wo rt hy of additi ona l c ommentis that est ab l i s he d in section 3.1.2 :t h eexistenc e of a ra di a l gradie nt in the m orpho logy of And I’s horiz o nt al branch .The se ns e o f t he gr a dien t is s uch that ther e are relativel y more blue ho r iz o n ta l bra nch stars outs idet he c orer ad ius . This grad ie n tc ould reflect either a n unde rlyin g abundance g radient or a radial a ge g r ad i ent, but given theabove disc u ssion, i t wou ld seemmore like l y that it sh oul d b e i n t er preted as sug g e stin gthat An d I was mo reext end edat its init ial star f or mi ng ep och,t han when t hebu lkof it s stars form ed. I no rde r to sh e dm o re l ig ht onthi spossi bili t y,we have sought t hee xist en ce of sim ilar gradient sin other d Sp h g alaxie s . This ta sk requires dSph studie s that f ulf ill t he f ollowingcri teria: a) the av ailabl e pho to met r y must e xt end b eyond thec o reradiu s; b)the pho tometry must be re a son ably preciseatthem a gn itu d eo f t he hor i z ontal branch; a nd c) theHB mo rphology m u stco ntain b oth blu e and red sta rs. These criteria a re s a t isf ied by thr ee studi es: thato f Car i na by S mec ker-Ha ne [* et al . *]{} ( 199 4), o f LeoII by De mers&Irwin (1 993), and of Sculptor b y Ligh t (19 88) . Consid eri n g f irst Cari na,the extens ive st udy o f S m ecker -Han e [ *et al. * ]{}( 1 994) clea r ly se p a ra tes the “re d c lum p” of in t ermedi ate- age core helium b u rning stars fr om t h e ho riz o ntal b ranch of old s tar s( b oth redan d blue). Sm ecker-Ha ne [*etal. *] {}(199 4) find n od iffere nceinthe surfa cede n sity di st ri b utions ofth ese tw o clas s es o f stars and furthe r, th e y indi c ate that t here is no s ystematictrend withradius inthe f raction o f thetot al core heli u m burning star popula ti on r epr esente d by t he ol d HB s tar s. Theser e su l ts a p ply out to a r adiu s of $\si m $10$\arc min $ beyond w hic h theirs am p l ing is inc omp lete. U sing

of_blue HB_and RR Lyrae variable_stars in_And_I is_that_they represent a_(small) population of_stars that formed perhaps_3 Gyr or_more_before the bulk of the stellar population in And I. Such a result implies_that,_just as_is_the_case for many of the_galactic dSphs, star formation in_And I_occurred over an extended interval. The driving mechanism(s)_behind_the extended and_diverse star formation histories of the Local Group dSph_systems remains one of the most_outstanding unsolved problems_in_this_branch of Astronomy. Horizontal Branch_Morphology Radial Gradients --------------------------------------------- A further property of_And I that is worthy of_additional comment is that established in section_3.1.2: the existence of a radial_gradient in the morphology of_And I’s_horizontal branch. The sense of_the gradient is_such that_there are relatively_more blue horizontal branch stars outside_the core radius._This gradient could reflect either an_underlying_abundance gradient or_a_radial_age gradient,_but given the_above_discussion, it_would_seem more likely that it should_be_interpreted as suggesting that And I was_more extended at its_initial_star forming epoch, than_when the bulk of its_stars formed. In order to shed more_light on_this possibility,_we have sought the existence of similar gradients in other dSph_galaxies. This task requires dSph studies_that fulfill the following_criteria: a)_the_available photometry must_extend_beyond the_core radius; b) the photometry must be_reasonably precise_at the magnitude of the horizontal_branch; and c) the_HB_morphology must contain both blue and_red stars. These criteria are satisfied_by three studies: that of_Carina_by_Smecker-Hane [*et al. *]{}(1994), of Leo_II by Demers & Irwin (1993),_and of Sculptor_by Light (1988). Considering first Carina, the extensive_study_of Smecker-Hane [*et al. *]{}(1994) clearly separates_the_“red clump” of intermediate-age core helium_burning_stars_from the horizontal branch of_old stars (both red and blue)._Smecker-Hane [*et al. *]{}(1994) find no difference in the surface_density distributions of_these two classes of stars_and_further,_they indicate that there is no systematic trend with radius_in the_fraction of the_total core helium burning star population represented by the old_HB stars. These results apply out to_a radius of $\sim$10$\arcmin$ beyond which their sampling is incomplete. Using

^0(J^{d_1} \times J^{d_2}))$. If $D=\sum_ia_i \pi^*(R_i), a_i \in \Bbb Z$, then we see that $\kappa_{\Bbb P_D}(\cal O_X(D)-\cal O_X)= \otimes_i (\cal P_{1R_i}^{\otimes 2}\otimes \cal P_{2R_i})^{\otimes a_i}$ and $\kappa_{\Bbb P_D}(\cal O_{C_0}(D)-\cal O_{C_0}) =\otimes_i \cal P_{1R_i}^{\otimes a_i}$. The assertion follows immediately from this. We shall first consider the case of $g \geq 2$. Assume that $g \geq 2$ and $H_x$ does not lie on a wall. \(1) If $r_1 \ne 1$, then $Pic(\overline{M}_{H_x}(r;c_1,c_2)) \cong Pic(J^{d_1} \times J^{d_2}) \oplus \Bbb Z^{\oplus 3}$. \(2) If $r_1=1$, then $$Pic(\overline{M}_{H_x}(r;c_1,c_2)) \cong \left\{ \begin{aligned} & Pic(J^{d_1} \times J^{d_2}) \oplus \Bbb Z^{\oplus 2}, \;x_1<x<x_0\\ & Pic(J^{d_1} \times J^{d_2}) \oplus \Bbb Z^{\oplus 3}, \;x<x_1. \end{aligned} \right.$$ \(3) $Pic(\overline{M}_{H_x}(r;c_1,c_2))/Pic(J^{d_1}\times J^{d_2})$ is generated by the image of $\kappa$. In the same way as in the proof of \[Y3, Lemma 4.2\], we see that ${\operatorname{codim}}(\Bbb P_i^{ss}) \geq 2$, ($i=1,2$) for a sufficiently large $d$. Then (1) follows immediately from Lemma \[lem:6\] and Lemma \[lem:14\]. We shall

^0(J^{d_1 } \times J^{d_2}))$. If $ D=\sum_ia_i \pi^*(R_i), a_i \in \Bbb Z$, then we see that $ \kappa_{\Bbb P_D}(\cal O_X(D)-\cal O_X)= \otimes_i (\cal P_{1R_i}^{\otimes 2}\otimes \cal P_{2R_i})^{\otimes a_i}$ and $ \kappa_{\Bbb P_D}(\cal O_{C_0}(D)-\cal O_{C_0 }) = \otimes_i \cal P_{1R_i}^{\otimes a_i}$. The assertion follows immediately from this. We shall first think the subject of $ g \geq 2$. Assume that $ g \geq 2 $ and $ H_x$ does not lie on a wall. \(1) If $ r_1 \ne 1 $, then $ Pic(\overline{M}_{H_x}(r;c_1,c_2) ) \cong Pic(J^{d_1 } \times J^{d_2 }) \oplus \Bbb Z^{\oplus 3}$. \(2) If $ r_1=1 $, then $ $ Pic(\overline{M}_{H_x}(r;c_1,c_2) ) \cong \left\ { \begin{aligned } & Pic(J^{d_1 } \times J^{d_2 }) \oplus \Bbb Z^{\oplus 2 }, \;x_1 < x < x_0\\ & Pic(J^{d_1 } \times J^{d_2 }) \oplus \Bbb Z^{\oplus 3 }, \;x < x_1. \end{aligned } \right.$$ \(3) $ Pic(\overline{M}_{H_x}(r;c_1,c_2))/Pic(J^{d_1}\times J^{d_2})$ is render by the image of $ \kappa$. In the same room as in the validation of \[Y3, Lemma 4.2\ ], we see that $ { \operatorname{codim}}(\Bbb P_i^{ss }) \geq 2 $, ($ i=1,2 $) for a sufficiently large $ d$. Then (1) follows immediately from Lemma \[lem:6\ ] and Lemma \[lem:14\ ]. We shall

^0(J^{d_1} \himes J^{d_2}))$. If $D=\sum_ia_i \pi^*(R_i), a_l \in \Bbb Z$, then cw see vhat $\kalpa_{\Bbb P_D}(\zal O_X(D)-\cal O_X)= \otimes_i (\cal P_{1R_i}^{\ltumes 2}\itimes \cal P_{2R_i})^{\otimes a_k}$ and $\kapia_{\Bbb P_D}(\cql O_{R_0}(D)-\cal O_{C_0}) =\otimes_i \rzl P_{1R_i}^{\obnmes z_l}$. The essertion folloes immediadely from this. Fe skall first consider the case of $g \gez 2$. Assumr hhat $g \geq 2$ anq $H_x$ qoes not lie on a wall. \(1) If $r_1 \ne 1$, then $Pjc(\overlpne{M}_{H_x}(r;c_1,c_2)) \cong Pic(K^{d_1} \times J^{d_2}) \oplus \Bbb Z^{\oplks 3}$. \(2) Lf $r_1=1$, then $$Pic(\overllne{M}_{H_x}(r;c_1,c_2)) \cing \lqdt\{ \begin{aligndd} & Pic(J^{d_1} \times J^{d_2}) \opmus \Bbb Z^{\oplus 2}, \;x_1<x<x_0\\ & Pic(J^{d_1} \times J^{d_2}) \o'lus \Bbb Z^{\oklhs 3}, \;x<x_1. \end{alignxd} \righn.$$ \(3) $Pic(\overline{M}_{H_x}(r;c_1,c_2))/Piw(J^{d_1}\timex J^{d_2})$ is generabed bb thw image of $\kappa$. In thx same way as in the proof of \[Y3, Lemma 4.2\], we see thqt ${\operdtortame{zidio}}(\Bbg '_i^{sa}) \geq 2$, ($i=1,2$) hor a suffidiently larte $d$. Then (1) follows omivciately from Lemma \[lqm:6\] and Lemma \[lem:14\]. We shall

^0(J^{d_1} \times J^{d_2}))$. If $D=\sum_ia_i \pi^*(R_i), a_i Z$, we see $\kappa_{\Bbb P_D}(\cal O_X(D)-\cal \cal a_i}$ and $\kappa_{\Bbb O_{C_0}(D)-\cal O_{C_0}) =\otimes_i P_{1R_i}^{\otimes a_i}$. The assertion follows immediately this. We shall first consider the case of $g \geq 2$. Assume that \geq 2$ and $H_x$ does not lie on a wall. \(1) If $r_1 1$, $Pic(\overline{M}_{H_x}(r;c_1,c_2)) Pic(J^{d_1} J^{d_2}) \oplus \Bbb Z^{\oplus 3}$. \(2) If $r_1=1$, then $$Pic(\overline{M}_{H_x}(r;c_1,c_2)) \cong \left\{ \begin{aligned} & Pic(J^{d_1} \times \oplus \Bbb Z^{\oplus 2}, \;x_1<x<x_0\\ & Pic(J^{d_1} \times \oplus \Bbb Z^{\oplus 3}, \end{aligned} \right.$$ \(3) $Pic(\overline{M}_{H_x}(r;c_1,c_2))/Pic(J^{d_1}\times J^{d_2})$ generated the image $\kappa$. the way as in proof of \[Y3, Lemma 4.2\], we see that ${\operatorname{codim}}(\Bbb P_i^{ss}) \geq 2$, ($i=1,2$) for a sufficiently large Then (1) from Lemma and \[lem:14\]. shall

^0(J^{d_1} \times J^{d_2}))$. If $D=\sum_ia_i \pi^*(R_i), a_i \iN \Bbb Z$, then wE see tHat $\KapPa_{\bbb P_d}(\cal o_X(D)-\cal O_X)= \otimes_I (\Cal P_{1r_i}^{\otimes 2}\otimes \cal P_{2R_i})^{\otImes a_I}$ aND $\kapPA_{\BBb P_D}(\cAl O_{C_0}(D)-\caL o_{C_0}) =\OTImeS_i \CaL P_{1R_I}^{\oTImEs a_i}$. THe aSsertioN follows imMedIaTely from this. wE sHall first cOnsIder the case oF $g \gEq 2$. AssuMe ThaT $G \geq 2$ aNd $H_X$ does Not lie ON a wall. \(1) if $r_1 \ne 1$, then $piC(\OverliNE{M}_{H_x}(r;c_1,c_2)) \CONg pic(J^{D_1} \times J^{d_2}) \oplus \Bbb Z^{\OPlUS 3}$. \(2) If $r_1=1$, then $$Pic(\oveRline{M}_{h_x}(R;C_1,c_2)) \CONg \lEft\{ \Begin{alignEd} & pic(J^{d_1} \TImes J^{d_2}) \oPLuS \bBB Z^{\oPLus 2}, \;x_1<x<x_0\\ & Pic(J^{d_1} \tiMes J^{d_2}) \oplus \BBB Z^{\oPlus 3}, \;x<x_1. \EnD{alIGned} \riGht.$$ \(3) $PiC(\oVErlIne{M}_{H_x}(r;c_1,c_2))/PiC(J^{d_1}\tImes J^{d_2})$ is gEneratED by the iMAge of $\kaPpa$. In tHe sAme Way aS In ThE prOoF Of \[Y3, lEmMa 4.2\], wE See That ${\operAtOrName{cOdim}}(\bBB p_I^{ss}) \gEq 2$, ($i=1,2$) For a SuffiCiently large $d$. theN (1) folLOws ImmedIatelY froM LEmma \[lEm:6\] and LEmma \[lEm:14\]. we shall

^0(J^{d_1} \times J^{d_2}) )$. If $D =\sum _ia _i\p i^*( R_i) , a_i \in \Bbb Z$,then we see that $\kap pa_{\ Bb b P_D } (\ cal O _X(D)-\ c al O _X) =\o tim es _ i(\cal P_ {1R_i}^ {\otimes 2 }\o ti mes \cal P_{ 2 R_ i})^{\otim esa_i}$ and $\ kap pa_{\B bb P_ D }(\ca l O _{C_0 }(D)-\ c al O_{ C_0}) =\o ti m es_i \ c al P_{1 R _ i} ^{\o times a_i}$. Thea ss e rtion followsimmedi at e ly f rom th is. We sh al l fir s t consi d er t h e c a se of $g \geq 2$. Assum e th at $g\g eq2 $ and$H_x$ d o esnot lie ona wa ll. \(1) If $r _ 1 \ne 1 $ , then$Pic(\ ove rli ne{M } _{ H_ x}( r; c _1, c _2 ))\ con g Pic(J^ {d _1 } \ti mesJ ^ { d _2}) \o plus \Bbb Z^{\oplus 3} $. \(2 ) If $r_1 =1$,then $ $Pic( \overl ine{M }_ {H_x}(r;c_1,c_2 )) \ cong \lef t\{ \ beg in {alig n ed} & Pi c(J ^{d_1}\timesJ ^{d _2 } ) \o plus \Bbb Z^{\oplu s2 } ,\;x_1<x< x_0\\&Pi c (J^{d_1} \ tim es J ^ { d_2}) \op l us \Bbb Z^ {\oplu s 3 }, \;x<x_ 1. \end{ al ign ed} \rig h t.$$ \(3) $Pic(\o verli n e{M}_{H_x}(r;c _ 1,c_2))/Pic(J ^ {d _ 1 }\ t imes J^ {d_2})$ isgene r ated byt he im a ge of $\ka pp a $. In the same way as i n theproof of \[Y3, Lem ma 4.2\],w e see that ${\ o pe r atorname{codim }}(\B bb P_i^{ss } ) \geq 2 $, ($ i=1,2$)for a suf f i cientlylar ge$d$ . T h e n(1) follows i m m edia te ly from Le mma \[l em: 6\] an d L em ma \[lem: 14\]. We s ha ll

^0(J^{d_1} \times J^{d_2}))$. If_$D=\sum_ia_i \pi^*(R_i),_a_i \in \Bbb Z$,_then we_see_that $\kappa_{\Bbb P_D}(\cal_O_X(D)-\cal_O_X)= \otimes_i (\cal P_{1R_i}^{\otimes_2}\otimes \cal P_{2R_i})^{\otimes_a_i}$ and $\kappa_{\Bbb P_D}(\cal_O_{C_0}(D)-\cal O_{C_0}) =\otimes_i \cal_P_{1R_i}^{\otimes_a_i}$. The assertion follows immediately from this. We shall first consider the case of $g_\geq_2$. Assume that_$g_\geq_2$ and $H_x$ does not_lie on a wall. \(1) If_$r_1 \ne_1$, then $Pic(\overline{M}_{H_x}(r;c_1,c_2)) \cong Pic(J^{d_1} \times J^{d_2}) \oplus \Bbb_Z^{\oplus_3}$. \(2) If $r_1=1$,_then $$Pic(\overline{M}_{H_x}(r;c_1,c_2)) \cong \left\{ \begin{aligned} & Pic(J^{d_1} \times J^{d_2}) \oplus_\Bbb Z^{\oplus 2}, \;x_1<x<x_0\\ & Pic(J^{d_1} \times_J^{d_2}) \oplus \Bbb_Z^{\oplus_3},_\;x<x_1. \end{aligned} \right.$$ \(3) $Pic(\overline{M}_{H_x}(r;c_1,c_2))/Pic(J^{d_1}\times J^{d_2})$ is_generated by the image of $\kappa$. In_the same way as in the_proof of \[Y3, Lemma 4.2\], we see_that ${\operatorname{codim}}(\Bbb P_i^{ss}) \geq 2$, ($i=1,2$)_for a sufficiently large $d$._Then (1)_follows immediately from Lemma \[lem:6\]_and Lemma \[lem:14\]._We shall

{\rm if $d =0$},$$ $$\sum_{\hat{0} \leq x \leq \hat{1}} B(\lbrack \hat{0}, x \rbrack; u,v) u^{d - \rho(x)} G(\lbrack x, \hat{1}\rbrack, u^{-1}v) = G(P,uv).$$ \[BfromG\] Let $P=[\hat 0,\hat 1]$ be an Eulerian poset. Then $$B(P;u,v) = \sum_{\hat 0\leq x \leq \hat 1} G([x,\hat 1]^*,u^{-1}v) (-u)^{\rk \hat 1 -\rk x} G([\hat 0,x],uv).$$ Indeed, one can sum the recursive formulas for $B([\hat 0, y])$ for all $\hat 0\leq y\leq\hat 1$ multiplied by $G([y,\hat 1]^*,u^{-1}v) (-u)^{\rk \hat 1 -\rk y}$ and use Lemma \[Ginverse\]. [BaBrFK]{} \[AKaMW\][AKMW]{} D. Abramovich, K. Karu, K. Matsuki, J. Włodarsczyk, [*Torification and Factorization of Birational Maps*]{}, preprint math.AG/9904135. \[BaBrFK\][bbfk]{} G. Barthel, J.-P. Brasselet, K.-H. Fieseler, L. Kaup, [*Combinatorial Intersection Cohomology for Fans*]{}, preprint math.AG/0002181. \[B1\][b1]{} V. V. Batyrev, [*Variations of the mixed Hodge structure of affine hypersurfaces in algebraic tori*]{}, Duke Math. J., [**69**]{} (1993), 349–409. \[B2\][b2]{}, [*Dual polyhedra and mirror symmetry for Calabi-Yau hypersurfaces in toric varieties*]{}, J. Algebraic Geometry [**6**]{} (1994), 493–535. \[B3\][Batyrev.cangor]{}, [*Stringy Hodge numbers of varieties with Gorenstein canonical singularities*]{}, Integrable systems and algebraic geometry (Kobe/Kyoto, 1997), 1–32, World Sci. Publishing, River Edge

{ \rm if $ d = 0$},$$ $ $ \sum_{\hat{0 } \leq x \leq \hat{1 } } B(\lbrack \hat{0 }, x \rbrack; u, v) u^{d - \rho(x) } G(\lbrack x, \hat{1}\rbrack, u^{-1}v) = G(P, uv).$$ \[BfromG\ ] Let $ P=[\hat 0,\hat 1]$ be an Eulerian poset. Then $ $ B(P;u, v) = \sum_{\hat 0\leq x \leq \hat 1 } G([x,\hat 1]^*,u^{-1}v) (-u)^{\rk \hat 1 -\rk x } G([\hat 0,x],uv).$$ Indeed, one can summarize the recursive formula for $ B([\hat 0, y])$ for all $ \hat 0\leq y\leq\hat 1 $ multiplied by $ G([y,\hat 1]^*,u^{-1}v) (-u)^{\rk \hat 1 -\rk y}$ and use Lemma \[Ginverse\ ]. [ BaBrFK ] { } \[AKaMW\][AKMW ] { } D. Abramovich, K. Karu, K. Matsuki, J. Włodarsczyk, [ * Torification and Factorization of Birational Maps * ] { }, preprint math. AG/9904135. \[BaBrFK\][bbfk ] { } G. Barthel, J.-P. Brasselet, K.-H. Fieseler, L. Kaup, [ * Combinatorial Intersection Cohomology for Fans * ] { }, preprint mathematics. AG/0002181. \[B1\][b1 ] { } V.   V. Batyrev, [ * Variations of the mixed Hodge structure of affine hypersurfaces in algebraic tori * ] { }, Duke Math. J., [ * * 69 * * ] { } (1993), 349–409. \[B2\][b2 ] { }, [ * double polyhedra and mirror symmetry for Calabi - Yau hypersurfaces in toric assortment * ] { }, J. Algebraic Geometry [ * * 6 * * ] { } (1994), 493–535. \[B3\][Batyrev.cangor ] { }, [ * Stringy Hodge numbers of varieties with Gorenstein canonic singularities * ] { }, Integrable systems and algebraic geometry (Kobe / Kyoto, 1997), 1–32, World Sci. Publishing, River Edge

{\rm lf $d =0$},$$ $$\sum_{\hat{0} \leq x \leq \hxt{1}} B(\lbrack \hat{0}, x \rbrack; u,v) u^{d - \rho(x)} G(\ubrack x, \hat{1}\rbrack, u^{-1}v) = G(P,uv).$$ \[UfronG\] Leu $P=[\hat 0,\hat 1]$ be an Ejlerian plset. Theb $$B(P;n,v) = \sum_{\hat 0\leq x \leq \hat 1} G([x,\haf 1]^*,u^{-1}v) (-u)^{\xk \hat 1 -\rk x} G([\hat 0,x],uv).$$ Indeed, one can sum tve rzcursive formulas for $B([\hat 0, y])$ for alj $\hat 0\lrq y\leq\hat 1$ multyplitd fy $G([g,\hat 1]^*,u^{-1}v) (-u)^{\rk \hat 1 -\rk y}$ and use Lemma \[Ginverve\]. [BaBrFK]{} \[AKaMW\][SKMW]{} D. Abramovich, K. Karu, K. Matduki, J. Włodarsczyk, [*Horificatiob anq Factorizatiun of Birauimnal Maps*]{}, kreprint math.AG/9904135. \[BaBrFK\][bbfk]{} G. Bartfel, J.-'. Brasselet, K.-Y. Flgseler, L. Kau', [*Combpnatorial Intcgsectiot Cohomplogy for Fans*]{}, prxpribt math.AG/0002181. \[B1\][b1]{} V. V. Batyret, [*Variations of the iixed Hodce structure of affunw hypgrsurxacer in alfeurajc torl*]{}, Dnke Math. J., [**69**]{} (1993), 349–409. \[B2\][b2]{}, [*Dual pilyhedra and mirror srnmetry for Camabi-Yat rypersurfaces in toric varieties*]{}, J. Algetrajc Geometry [**6**]{} (1994), 493–535. \[B3\][Batyrev.cqngor]{}, [*Stringy Hodge nombers of darieties with Gorenstein canonical singularities*]{}, Intejrxblt wysteow wnd algebraic geometry (Kobe/Kyoto, 1997), 1–32, World Sci. [hbkixhing, River Edne

{\rm if $d =0$},$$ $$\sum_{\hat{0} \leq x B(\lbrack x \rbrack; u^{d - \rho(x)} G(P,uv).$$ Let $P=[\hat 0,\hat be an Eulerian Then $$B(P;u,v) = \sum_{\hat 0\leq x \hat 1} G([x,\hat 1]^*,u^{-1}v) (-u)^{\rk \hat 1 -\rk x} G([\hat 0,x],uv).$$ Indeed, one sum the recursive formulas for $B([\hat 0, y])$ for all $\hat 0\leq y\leq\hat multiplied $G([y,\hat (-u)^{\rk 1 -\rk y}$ and use Lemma \[Ginverse\]. [BaBrFK]{} \[AKaMW\][AKMW]{} D. Abramovich, K. Karu, K. Matsuki, J. [*Torification and Factorization of Birational Maps*]{}, preprint math.AG/9904135. G. Barthel, J.-P. Brasselet, Fieseler, L. Kaup, [*Combinatorial Intersection for preprint math.AG/0002181. V. Batyrev, of the mixed structure of affine hypersurfaces in algebraic tori*]{}, Duke Math. J., [**69**]{} (1993), 349–409. \[B2\][b2]{}, [*Dual polyhedra and symmetry for in toric J. Geometry (1994), 493–535. \[B3\][Batyrev.cangor]{}, numbers of varieties with Gorenstein canonical and algebraic geometry (Kobe/Kyoto, 1997), 1–32, World Sci. River Edge

{\rm if $d =0$},$$ $$\sum_{\hat{0} \leq x \leq \hat{1}} B(\lbrAck \hat{0}, x \rbrAck; u,v) U^{d - \rHo(x)} g(\lBracK x, \haT{1}\rbrack, u^{-1}v) = G(P,uv).$$ \[BFRomG\] let $P=[\hat 0,\hat 1]$ be an Eulerian Poset. thEN $$B(P;u,V) = \SuM_{\hat 0\lEq x \leq \hAT 1} G([X,\HAt 1]^*,u^{-1}V) (-u)^{\Rk \Hat 1 -\Rk X} g([\hAt 0,x],uv).$$ indEed, one cAn sum the reCurSiVe formulas foR $b([\hAt 0, y])$ for all $\hAt 0\lEq y\leq\hat 1$ mulTipLied by $g([y,\Hat 1]^*,U^{-1}V) (-u)^{\rk \hAt 1 -\rK y}$ and Use LemMA \[GinveRse\]. [BaBrFK]{} \[aKAmW\][AKMW]{} d. abramovICH, K. karu, k. Matsuki, J. WłodarscZYk, [*tOrification and factorIzATiON Of BIraTional Maps*]{}, PrEprinT Math.AG/9904135. \[BAbrfk\][BBfk]{} g. barthel, J.-P. BrasSelet, K.-H. FiesELer, l. Kaup, [*COmBinATorial interSeCTioN Cohomology For FAns*]{}, prepriNt math.ag/0002181. \[B1\][b1]{} V. V. BaTYrev, [*VarIationS of The MixeD hoDgE stRuCTurE Of AffINe hYpersurfAcEs In algEbraIC TORi*]{}, DuKe MAth. J., [**69**]{} (1993), 349–409. \[b2\][b2]{}, [*DuaL polyhedra and MirRor sYMmeTry foR CalaBi-YaU hYpersUrfaceS in toRiC varieties*]{}, J. AlgeBraiC Geometry [**6**]{} (1994), 493–535. \[b3\][BaTyRev.CaNgor]{}, [*STRingy HOdgE nuMbers of VarietiES wiTh gOREnStein canonical singUlARItIes*]{}, IntegRable sYStEmS And algebRaIc gEomeTRY (Kobe/kyotO, 1997), 1–32, woRld Sci. PuBlishiNG, RIvEr Edge

{\rm if $d =0$},$$ $$\sum_ {\hat{0} \ leq x \l eq\h at{1 }} B (\lbrack \hat{ 0 }, x \rbrack; u,v) u^{d - \rho (x ) } G( \ lb rackx, \ha t {1 } \ rbr ac k, u^ {- 1 }v ) = G (P, uv).$$ \[BfromG\ ] L et $P=[\hat 0, \ ha t 1]$ be a n E ulerian pose t.Then $ $B (P; u ,v) = \s um_{\ hat 0\ l eq x \ leq \hat1} G([x,\ h at 1]^* , u ^{ -1}v ) (-u)^{\rk \hat1 - \ rk x} G([\hat0,x],u v) . $$ Ind eed , one cansu m the recursi v ef o r mul a s for $B([\ha t 0, y])$ f o r a ll $\h at 0\ l eq y\l eq\ha t1 $ m ultiplied b y $G ([y,\hat1]^*,u ^ {-1}v)( -u)^{\r k \hat 1-\r k y} $ a nd us eL emm a \ [Gi n ver se\]. [ Ba Br FK]{} \[ A K a M W\][ AKM W]{} D. A bramovich, K. Ka ru,K . M atsuk i, J. Wło da rsczy k, [*T orifi ca tion and Factor izat ion of Bi rat io nal M aps*] { }, pre pri ntmath.AG /990413 5 . \[ B a B rF K\][bbfk]{} G. Bar th e l ,J.-P. Br assele t ,K. - H. Fiese le r,L. K a u p, [* Comb i na torial I nterse c ti on Cohomo lo gy for F ans *]{ }, pr e prin t math .AG/0002 181.\[B1\][b1]{} V .  V. Batyrev,[ *V a r ia t ions of the mixedHodg e str uctu r eofa ffine hype rs u rf a ces in algebraic to ri *]{},DukeMath. J., [** 69**]{} (1 9 9 3 ), 349–4 09.\[ B 2\][b2]{}, [*D ual p olyhedra a n d mirror symm etry for Calabi-Y a u hypersu rfa ces in to r i cvarieties*]{} , J. A lg ebraicGeo metry [ **6 **] {}(19 94 ), 493–53 5. \[B3 \] [B at yr ev. cango r ]{}, [*S tr ing yHod ge nu m bers o f var ieti es w i thGorenst e in c anon ic al sin gul ar ities *]{} , In tegrabl e systems an d alg eb ra ic geom etry (Kobe/Ky ot o, 1997),1– 32, World S ci. Publ ishing, River Edge

{\rm if_$d =0$},$$_$$\sum_{\hat{0} \leq x \leq_\hat{1}} B(\lbrack \hat{0},_x_\rbrack; u,v)_u^{d_ - \rho(x)} G(\lbrack_x, \hat{1}\rbrack,_u^{-1}v) = G(P,uv).$$ \[BfromG\] Let_$P=[\hat 0,\hat 1]$_be_an Eulerian poset. Then $$B(P;u,v) = \sum_{\hat 0\leq x \leq \hat 1} G([x,\hat 1]^*,u^{-1}v) (-u)^{\rk_\hat_1 -\rk_x} G([\hat_0,x],uv).$$ Indeed,_one can sum the recursive_formulas for $B([\hat 0, y])$_for all_$\hat 0\leq y\leq\hat 1$ multiplied by $G([y,\hat 1]^*,u^{-1}v) (-u)^{\rk_\hat_1 -\rk y}$_and use Lemma \[Ginverse\]. [BaBrFK]{} \[AKaMW\][AKMW]{} D. Abramovich, K. Karu, K._Matsuki, J. Włodarsczyk, [*Torification and Factorization_of Birational Maps*]{},_preprint_math.AG/9904135. \[BaBrFK\][bbfk]{}_G. Barthel, J.-P. Brasselet,_K.-H. Fieseler, L. Kaup, [*Combinatorial Intersection_Cohomology for Fans*]{}, preprint math.AG/0002181. \[B1\][b1]{} V. V._Batyrev, [*Variations of the mixed Hodge structure_of affine hypersurfaces in algebraic tori*]{},_Duke Math. J., [**69**]{} (1993),_349–409. \[B2\][b2]{}, [*Dual_polyhedra and mirror symmetry for_Calabi-Yau hypersurfaces in_toric varieties*]{},_J. Algebraic Geometry_[**6**]{} (1994), 493–535. \[B3\][Batyrev.cangor]{}, [*Stringy Hodge numbers_of varieties with_Gorenstein canonical singularities*]{}, Integrable systems and_algebraic_geometry (Kobe/Kyoto, 1997),_1–32,_World_Sci. Publishing,_River Edge

},\mathbf{u}^{p}_{0:t-1})\mathbf{H}^{p}_t(\mathbf{x}^{p}_{t}))\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{x}^{p}_{0:t},\mathbf{u}^{p}_{0:t-1})\end{aligned}$$ which is solvable with the initial condition $ \mathbf{P}^{+}_{\mathbf{b}^{p}_0}(\mathbf{x}^{p}_{0})=\boldsymbol{\Sigma}_{\mathbf{x}_{0}} $. \[problem:Planning Problem\] Given an initial belief $ \mathbf{b}_{0}\in\mathbb{B} $, a goal region represented as an $ \ell_2 $-norm ball, $ \mathbf{B}_{r_g}(\mathbf{x}_g) $, of radius $ r_g $ around a goal state $ \mathbf{x}_{g}\in\mathbb{X} $, and a planning horizon of $ K>0 $, we define the following problem: $$\begin{aligned} \nonumber \min_{\mathbf{u}^{p}_{0:K-1}}\sum\limits_{t=1}^{K}[&\mathrm{tr}(\mathbf{W}_t \mathbf{P}^{+}_{\mathbf{b}^{p}_t}\mathbf{W}^T_t)+({\mathbf{u}}^{p}_{t-1})^{T}\mathbf{W}^{u}_{t}{\mathbf{u}}^{p}_{t-1}] \\s.t.~~\mathbf{P}^{-}_{\mathbf{b}^{p}_t}&=\mathbf{A}^{p}_t\mathbf{P}^{+}_{\mathbf{b}^{p}_{t-1}}(\mathbf{A}^{p}_t)^T+\mathbf{G}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\omega}_{t}}(\mathbf{G}_t)^T\label{eq:riccati 1} \\\mathbf{S}_{\mathbf{b}^{p}_t}&=\mathbf{H}^{p}_t\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{H}^{p}_t)^T+\mathbf{M}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\nu}_{t}}(\mathbf{M}^{p}_t)^T\label{eq:riccati 2} \\\mathbf{K}_{\mathbf{b}^{p}_t}&=\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\

}, \mathbf{u}^{p}_{0: t-1})\mathbf{H}^{p}_t(\mathbf{x}^{p}_{t}))\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{x}^{p}_{0: t},\mathbf{u}^{p}_{0: t-1})\end{aligned}$$ which is solvable with the initial condition $ \mathbf{P}^{+}_{\mathbf{b}^{p}_0}(\mathbf{x}^{p}_{0})=\boldsymbol{\Sigma}_{\mathbf{x}_{0 } } $. \[problem: Planning Problem\ ] Given an initial belief $ \mathbf{b}_{0}\in\mathbb{B } $, a goal area stage as an $ \ell_2 $ -norm ball, $ \mathbf{B}_{r_g}(\mathbf{x}_g) $, of radius $ r_g $ around a finish state $ \mathbf{x}_{g}\in\mathbb{X } $, and a planning horizon of $ K>0 $, we specify the following problem: $ $ \begin{aligned } \nonumber \min_{\mathbf{u}^{p}_{0: K-1}}\sum\limits_{t=1}^{K}[&\mathrm{tr}(\mathbf{W}_t \mathbf{P}^{+}_{\mathbf{b}^{p}_t}\mathbf{W}^T_t)+({\mathbf{u}}^{p}_{t-1})^{T}\mathbf{W}^{u}_{t}{\mathbf{u}}^{p}_{t-1 } ] \\s.t.~~\mathbf{P}^{-}_{\mathbf{b}^{p}_t}&=\mathbf{A}^{p}_t\mathbf{P}^{+}_{\mathbf{b}^{p}_{t-1}}(\mathbf{A}^{p}_t)^T+\mathbf{G}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\omega}_{t}}(\mathbf{G}_t)^T\label{eq: riccati 1 } \\\mathbf{S}_{\mathbf{b}^{p}_t}&=\mathbf{H}^{p}_t\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{H}^{p}_t)^T+\mathbf{M}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\nu}_{t}}(\mathbf{M}^{p}_t)^T\label{eq: riccati 2 } \\\mathbf{K}_{\mathbf{b}^{p}_t}&=\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\

},\matjbf{u}^{p}_{0:t-1})\mathbf{H}^{p}_t(\mathbf{x}^{p}_{t}))\oathbf{P}^{-}_{\mathbf{b}^{p}_j}(\mqthbf{x}^{'}_{0:t},\mathbr{u}^{p}_{0:t-1})\end{auigned}$$ which is solvable witi thw iniupal condition $ \mathbf{O}^{+}_{\mathbf{b}^{p}_0}(\lathbf{x}^{p}_{0})=\voldwtmbol{\Sigma}_{\mathbf{x}_{0}} $. \[ixoblej:Ilannnnj Problem\] Given an initian belief $ \mathtf{c}_{0}\iu\mathbb{B} $, a goal region represented ws an $ \rlp_2 $-norm ball, $ \mwthbg{F}_{r_g}(\mznhnf{x}_g) $, of radius $ r_g $ around a gozl statt $ \mathbf{x}_{g}\in\mathbn{X} $, and a planning horizon of $ K>0 $, we define the vollowing ptkblqn: $$\begin{alignea} \nonumber \min_{\mathbf{u}^{p}_{0:K-1}}\aum\limits_{t=1}^{K}[&\mathrm{tr}(\mathbf{W}_t \matfbf{P}^{+}_{\mcthbf{b}^{p}_t}\matybd{W}^T_h)+({\kathbf{u}}^{p}_{t-1})^{T}\methbf{W}^{l}_{t}{\mathbf{u}}^{p}_{t-1}] \\s.t.~~\mathbf{P}^{-}_{\mdthbf{b}^{p}_y}&=\mathbf{A}^{p}_t\mathnf{P}^{+}_{\mavhbf{v}^{p}_{t-1}}(\mathbf{A}^{p}_t)^T+\mathbf{G}^{p}_v\boldsymbol{\Sigma}_{\boldfymbol{\omeca}_{c}}(\mathbf{G}_t)^T\label{eq:riccqtu 1} \\\matvbf{S}_{\kathcd{b}^{p}_g}&=\mauhbh{H}^{p}_f\mathbv{P}^{-}_{\methbf{b}^{p}_t}(\matgbf{H}^{p}_t)^T+\mathvf{M}^{p}_t\boldsymbol{\Sigms}_{\bjodsymbol{\nu}_{t}}(\mafhbf{M}^{p}_e)^T\jabel{eq:riccati 2} \\\mathbf{K}_{\mathbf{b}^{p}_t}&=\mathbf{P}^{-}_{\mdthgf{b}^{p}_t}(\

},\mathbf{u}^{p}_{0:t-1})\mathbf{H}^{p}_t(\mathbf{x}^{p}_{t}))\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{x}^{p}_{0:t},\mathbf{u}^{p}_{0:t-1})\end{aligned}$$ which is solvable with the initial \mathbf{P}^{+}_{\mathbf{b}^{p}_0}(\mathbf{x}^{p}_{0})=\boldsymbol{\Sigma}_{\mathbf{x}_{0}} \[problem:Planning Problem\] an initial belief region as an $ $-norm ball, $ $, of radius $ r_g $ a goal state $ \mathbf{x}_{g}\in\mathbb{X} $, and a planning horizon of $ K>0 we define the following problem: $$\begin{aligned} \nonumber \min_{\mathbf{u}^{p}_{0:K-1}}\sum\limits_{t=1}^{K}[&\mathrm{tr}(\mathbf{W}_t \mathbf{P}^{+}_{\mathbf{b}^{p}_t}\mathbf{W}^T_t)+({\mathbf{u}}^{p}_{t-1})^{T}\mathbf{W}^{u}_{t}{\mathbf{u}}^{p}_{t-1}] \\s.t.~~\mathbf{P}^{-}_{\mathbf{b}^{p}_t}&=\mathbf{A}^{p}_t\mathbf{P}^{+}_{\mathbf{b}^{p}_{t-1}}(\mathbf{A}^{p}_t)^T+\mathbf{G}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\omega}_{t}}(\mathbf{G}_t)^T\label{eq:riccati 1} \\\mathbf{S}_{\mathbf{b}^{p}_t}&=\mathbf{H}^{p}_t\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{H}^{p}_t)^T+\mathbf{M}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\nu}_{t}}(\mathbf{M}^{p}_t)^T\label{eq:riccati 2}

},\mathbf{u}^{p}_{0:t-1})\mathbf{H}^{p}_t(\mathbf{x}^{P}_{t}))\mathbf{P}^{-}_{\mAthbf{B}^{p}_t}(\MatHbF{x}^{p}_{0:t},\MathBf{u}^{p}_{0:t-1})\end{aligneD}$$ WhicH is solvable with the initIal coNdITion $ \MAtHbf{P}^{+}_{\mAthbf{b}^{p}_0}(\MAtHBF{x}^{p}_{0})=\BoLdSymBoL{\siGma}_{\maThbF{x}_{0}} $. \[problEm:Planning proBlEm\] Given an iniTIaL belief $ \matHbf{B}_{0}\in\mathbb{B} $, a gOal Region RePreSEnted As aN $ \ell_2 $-nOrm balL, $ \Mathbf{b}_{r_g}(\mathbf{X}_g) $, OF radiuS $ R_g $ arounD A GoAl stAte $ \mathbf{x}_{g}\in\mathBB{X} $, ANd a planning horIzon of $ k>0 $, wE DeFINe tHe fOllowing prObLem: $$\beGIn{alignED} \nONUMbeR \Min_{\mathbf{u}^{p}_{0:K-1}}\sUm\limits_{t=1}^{K}[&\mAThrM{tr}(\matHbF{W}_t \MAthbf{P}^{+}_{\MathbF{b}^{P}_T}\maThbf{W}^T_t)+({\mathBf{u}}^{p}_{T-1})^{T}\mathbf{W}^{U}_{t}{\mathBF{u}}^{p}_{t-1}] \\s.t.~~\mAThbf{P}^{-}_{\maThbf{b}^{p}_T}&=\maThbF{A}^{p}_t\MAtHbF{P}^{+}_{\mAtHBf{b}^{P}_{T-1}}(\mAthBF{A}^{p}_T)^T+\mathbf{g}^{p}_T\bOldsyMbol{\sIGMA}_{\bolDsyMbol{\Omega}_{T}}(\mathbf{G}_t)^T\labEl{eQ:ricCAti 1} \\\MathbF{S}_{\matHbf{b}^{P}_t}&=\MathbF{H}^{p}_t\maThbf{P}^{-}_{\MaThbf{b}^{p}_t}(\mathbf{H}^{p}_T)^T+\maThbf{M}^{p}_t\boLdsYmBol{\siGma}_{\boLDsymboL{\nu}_{T}}(\maThbf{M}^{p}_t)^t\label{eQ:RicCaTI 2} \\\MAtHbf{K}_{\mathbf{b}^{p}_t}&=\mathbF{P}^{-}_{\MAThBf{b}^{p}_t}(\

},\mathbf{u}^{p}_{0:t-1})\ mathbf{H}^ {p}_t (\m ath bf {x}^ {p}_ {t}))\mathbf{P } ^{-} _{\mathbf{b}^{p}_t}(\m athbf {x } ^{p} _ {0 :t},\ mathbf{ u }^ { p }_{ 0: t- 1}) \e n d{ align ed} $$ whic h is solva ble w ith the init i al condition $\mathbf{P}^{ +}_ {\math bf {b} ^ {p}_0 }(\ mathb f{x}^{ p }_{0}) =\boldsym bo l {\Sigm a }_{\mat h b f{ x}_{ 0}} $. \[problem : Pl a nning Problem\ ] Give na ni n iti albelief $ \ ma thbf{ b }_{0}\i n \m a t h bb{ B } $, a goal r egion repre s ent ed asan $\ ell_2$-nor mb all , $ \mathbf {B}_ {r_g}(\ma thbf{x } _g) $,o f radiu s $ r_ g $ ar ound ago alst a te$ \ mat h bf{ x}_{g}\i n\ ma thbb{ X} $ , a n d apla nnin g hor izon of $ K>0 $, wed efi ne th e fol lowi ng prob lem: $$\be gi n{aligned} \non umbe r \min_{\ mat hb f{u }^ {p}_{ 0 :K-1}} \su m\l imits_{ t=1}^{K } [&\ ma t h r m{ tr}(\mathbf{W}_t \ ma t h bf {P}^{+}_ {\math b f{ b} ^ {p}_t}\m at hbf {W}^ T _ t)+({ \mat h bf {u}}^{p} _{t-1} ) ^{ T} \mathbf {W }^{u}_ {t }{\ mat hbf{u } }^{p }_{t-1 }] \\s.t .~~\m a thbf{P}^{-}_{\ m athbf{b}^{p}_ t }& = \ ma t hbf{ A}^ {p}_t\mathb f{P} ^ {+}_ {\ma t hb f{b } ^{p}_ {t-1} }( \ ma t hbf{A}^{p}_t)^T+\ma th bf{G}^ {p}_t \boldsymbol{\ Sigma}_{\b o l d symbol{\ omeg a }_ { t}}(\mathbf{G} _t)^T \label{eq: r iccati 1 } \\\ mathbf{S }_{\mathb f { b}^{p}_t }&= \ma thb f{H } ^ {p }_t\mathbf{P} ^ { -}_{ \m athbf{b }^{ p}_t}(\ mat hbf {H} ^{p }_ t)^T+\mat hbf{M}^{ p} _t \b ol dsy mbol{ \ Sigma}_{ \b old sy mbo l{\nu } _{t}}( \math bf{M }^ {p } _t) ^T\labe l {e q : ricc at i2} \ \\m at hbf{K }_{\ m ath bf{b}^{ p}_t}&=\m ath b f{P} ^{ -} _{\math bf{b}^{p}_t}( \

},\mathbf{u}^{p}_{0:t-1})\mathbf{H}^{p}_t(\mathbf{x}^{p}_{t}))\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{x}^{p}_{0:t},\mathbf{u}^{p}_{0:t-1})\end{aligned}$$ which_is solvable_with the initial condition_$ \mathbf{P}^{+}_{\mathbf{b}^{p}_0}(\mathbf{x}^{p}_{0})=\boldsymbol{\Sigma}_{\mathbf{x}_{0}}_$. \[problem:Planning_Problem\] Given_an_initial belief $_\mathbf{b}_{0}\in\mathbb{B} $, a_goal region represented as_an $ \ell_2_$-norm_ball, $ \mathbf{B}_{r_g}(\mathbf{x}_g) $, of radius $ r_g $ around a goal state $_\mathbf{x}_{g}\in\mathbb{X}_$, and_a_planning_horizon of $ K>0 $,_we define the following problem: $$\begin{aligned} \nonumber_\min_{\mathbf{u}^{p}_{0:K-1}}\sum\limits_{t=1}^{K}[&\mathrm{tr}(\mathbf{W}_t \mathbf{P}^{+}_{\mathbf{b}^{p}_t}\mathbf{W}^T_t)+({\mathbf{u}}^{p}_{t-1})^{T}\mathbf{W}^{u}_{t}{\mathbf{u}}^{p}_{t-1}] \\s.t.~~\mathbf{P}^{-}_{\mathbf{b}^{p}_t}&=\mathbf{A}^{p}_t\mathbf{P}^{+}_{\mathbf{b}^{p}_{t-1}}(\mathbf{A}^{p}_t)^T+\mathbf{G}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\omega}_{t}}(\mathbf{G}_t)^T\label{eq:riccati_1} \\\mathbf{S}_{\mathbf{b}^{p}_t}&=\mathbf{H}^{p}_t\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\mathbf{H}^{p}_t)^T+\mathbf{M}^{p}_t\boldsymbol{\Sigma}_{\boldsymbol{\nu}_{t}}(\mathbf{M}^{p}_t)^T\label{eq:riccati 2} \\\mathbf{K}_{\mathbf{b}^{p}_t}&=\mathbf{P}^{-}_{\mathbf{b}^{p}_t}(\

I. A. Lubashevsky, A. A. Zemlyanov, Continuum description of anomalous diffusion on a comb structure, J. Exp. Theor. Phys. **87** (1998) 700–713. C. W. Gardiner, Handbook of Stochastic Methods: for Physics, Chemistry and the Natural Sciences, Springer-Verlag, Berlin, 2004, 3-rd edition. --- abstract: 'A virtual graphical construction is made to show the difference between neutrino and anti-neutrino oscillations in the presence of CP violation with CPT conservation.' author: - | R. G. Moorhouse$^b$\ $^b$University of Glasgow,Glasgow G12 8QQ, U.K.\ title: Note on a Pattern from CP Violation in Neutrino Oscillations --- Introduction ============ There is interest in the possibility that CPT violation may occur and then show in neutrino oscillation experiments[@BL],[@CDK], [@MINOS].. However this may be, CP violation is long established and it is of importance to seek it in neutrino oscillation results. If one takes a conservative point of view that nature conserves CPT, and there are 3 generations of neutrinos, then the consequences of CP violation in neutrino oscillation become more definite. In particular if CP were conserved $\nu$ transition probabilities would be the same as $\bar{\nu}$ transition probabilities while the occurence therein of CP violation makes these different [@Kayser]. This differencc is dependent on the neutrino oscillation parameter $L/E$, $L$ being the distance of travel from creation to detection and $E$ the energy of the initial neutrino: and it is also sensitive to the value of the small ratio of neutrino mass squared differences. There results a complicated 2-variable dependence in addition to the linear dependence on the leptonic Jarlskog parameter, $J_{lep}$[@HSW]. Pattern for the difference between $\nu$ and $\bar\nu$ oscillations =================================================================== The input to the formula for neutrino transition probabilities is largely from the mixing matrix elements $U_{\alpha i}$ where $\alpha$ is one of the 3 flavour indices and $i$ one of the 3 mass eigenstate indices. From these 9 elements plaquettes [@BD] can be constructed, these being phase invariant products of 2 $U$ elements multiplied by products of 2 $U^{\star

I. A. Lubashevsky, A. A. Zemlyanov, Continuum description of anomalous diffusion on a comb structure, J. Exp. Theor. Phys. * * 87 * * (1998) 700–713. C. W. Gardiner, Handbook of Stochastic Methods: for Physics, Chemistry and the Natural Sciences, Springer - Verlag, Berlin, 2004, 3 - rd edition. --- abstract:' A virtual graphic structure is made to show the remainder between neutrino and anti - neutrino oscillation in the presence of CP violation with CPT conservation.' generator: - | R.   G.   Moorhouse$^b$\ $ ^b$University of Glasgow, Glasgow G12 8QQ, U.K.\ deed: Note on a Pattern from CP Violation in Neutrino Oscillations --- Introduction = = = = = = = = = = = = There is interest in the hypothesis that CPT violation may occur and then show in neutrino oscillation experiments[@BL],[@CDK ], [ @MINOS ].. However this may be, CP irreverence is long established and it is of importance to seek it in neutrino oscillation results. If one takes a cautious point of view that nature conserves CPT, and there exist 3 generations of neutrinos, then the consequences of CP violation in neutrino cycle become more definite. In particular if CP were conserve $ \nu$ transition probabilities would be the same as $ \bar{\nu}$ passage probabilities while the occurence therein of CP violation makes these different [ @Kayser ]. This differencc is dependent on the neutrino oscillation parameter $ L / E$, $ L$ being the distance of travel from creation to detection and $ E$ the energy of the initial neutrino: and it is also sensitive to the value of the small ratio of neutrino mass square differences. There results a complicated 2 - variable star addiction in addition to the linear dependence on the leptonic Jarlskog parameter, $ J_{lep}$[@HSW ]. Pattern for the remainder between $ \nu$ and $ \bar\nu$ oscillations = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = The input to the formula for neutrino transition probability is largely from the mixing matrix elements $ U_{\alpha i}$ where $ \alpha$ is one of the 3 flavour indices and $ i$ one of the 3 mass eigenstate indices. From these 9 element plaquettes [ @BD ] can be constructed, these being phase invariant products of 2 $ U$ elements breed by products of 2 $ U^{\star

I. A. Lubashevsky, A. A. Zemlyanuv, Continuum descriptimn of znomalour diffusion on a comb structnre, H. Exp. Theor. Phys. **87** (1998) 700–713. C. W. Gardkner, Handhook of Wtociastic Methods: for Physigf, Chslistxy and the Natursl Sciencev, Springer-Verldg, Bzrlin, 2004, 3-rd edition. --- abstract: 'A virtual graphivap construction is kwde fo show the difference between neufrino aid anti-neutrino oscillations in the presejce lf CP violation wihh CPT consgdvaeuon.' author: - | R. G. Moorholve$^b$\ $^b$Unjversity of Glasgow,Glasgow G12 8QQ, U.K.\ ticle: Note on a Pahjern from CP Violwtion in Neubgino Oswillatipns --- Introductiok ============ Therx is interest in the possmbility that CPT viojation maf kccur and then shiw in ngutrito orxiluatjoi esperimfnta[@BL],[@CDK], [@MINKS].. However rhis may be, CP violstyin is long esfablisreq and it is of importance to seek it in nehtrino oscillation resuots. If one takes a cojservativq point of view that nature conserves CPT, and these arx 3 geucvatiubs of neutrinos, then the consequences of CP viojztook in neutrino osgillation become mprf cgfinite. In parjicular if CP were conserved $\nu$ tragsitiin probabylitoes would be the same as $\bae{\nu}$ transitijb probabilities whnle the occuxence jhereim of CP violation makes theae differenh [@Kayser]. Ffis differencc ir dvpengent on ufe neutrino osciljation pacametzr $L/E$, $L$ ceinb the qistance ov trayal from creation tl detgction and $E$ the energy of the initial neutrino: and it is alsp venvitive tj the value of the small ratio oy neutriuo masr squared siffereices. There rqsults a compnlcated 2-variaule depenqencw in additiuv to the lineat dependeuee on the leptonic Jarlskog patajeter, $J_{lep}$[@HSW]. Pacuerb for the diffetenze fenwexn $\nu$ dnd $\bar\nu$ oswillxtiums =================================================================== The input uj bhe forkula for neutrino trdnsifion probabilities if largeli from thq mixing matrox elements $U_{\alpha i}$ whxre $\al'ha$ is ong of the 3 flavour indices and $i$ one of tje 3 mass eigenseate indices. Frok these 9 elements plaquettes [@BD] can be ronstructed, these being phase invariant prpcucts of 2 $U$ elemqnts multhplied by products od 2 $U^{\star

I. A. Lubashevsky, A. A. Zemlyanov, Continuum anomalous on a structure, J. Exp. C. Gardiner, Handbook of Methods: for Physics, and the Natural Sciences, Springer-Verlag, Berlin, 3-rd edition. --- abstract: 'A virtual graphical construction is made to show the between neutrino and anti-neutrino oscillations in the presence of CP violation with CPT author: | G. $^b$University of Glasgow,Glasgow G12 8QQ, U.K.\ title: Note on a Pattern from CP Violation in Neutrino --- Introduction ============ There is interest in the that CPT violation may and then show in neutrino experiments[@BL],[@CDK], However this be, violation long established and is of importance to seek it in neutrino oscillation results. If one takes a conservative point of that nature and there 3 of then the consequences violation in neutrino oscillation become more if CP were conserved $\nu$ transition probabilities would the same $\bar{\nu}$ transition probabilities while the occurence of CP violation makes these different [@Kayser]. This is dependent on the neutrino oscillation parameter $L/E$, $L$ being the distance of travel from detection and $E$ the of the initial and is sensitive the value the small ratio of neutrino mass squared differences. There results a 2-variable dependence in addition to the linear dependence on the parameter, Pattern for the between $\nu$ and $\bar\nu$ =================================================================== input to the formula transition is mixing elements i}$ where $\alpha$ is of the 3 flavour indices $i$ one of the these 9 elements plaquettes [@BD] can be constructed, being phase invariant products of 2 $U$ multiplied by products of 2 $U^{\star

I. A. Lubashevsky, A. A. Zemlyanov, COntinuum deScripTioN of AnOmalOus dIffusion on a comB StruCture, J. Exp. Theor. Phys. **87** (1998) 700–713. C. W. GaRdineR, HANdboOK oF StocHastic METhODS: foR PHySicS, CHEmIstry And The NatuRal ScienceS, SpRiNger-Verlag, BeRLiN, 2004, 3-rd edition. --- AbsTract: 'A virtuaL grAphicaL cOnsTRuctiOn iS made To show THe diffErence betWeEN neutrINo and anTI-NeUtriNo oscillations in tHE pREsence of CP violAtion wItH cPt COnsErvAtion.' authoR: - | R. g. MoorHOuse$^b$\ $^b$UNIvERSIty OF Glasgow,GlasgOw G12 8QQ, U.K.\ titlE: notE on a PaTtErn FRom CP VIolatIoN In NEutrino OsciLlatIons --- IntroDuctioN ============ there is INterest In the pOssIbiLity THaT CpT vIoLAtiON mAy oCCur And then sHoW iN neutRino OSCILlatIon ExpeRimenTs[@BL],[@CDK], [@MINOS].. HOweVer tHIs mAy be, Cp violAtioN iS long EstablIshed AnD it is of importanCe to Seek it in nEutRiNo oScIllatIOn resuLts. if oNe takes A conserVAtiVe POINt Of view that nature coNsERVeS CPT, and tHere arE 3 GeNeRAtions of NeUtrInos, THEn the ConsEQuEnces of Cp violaTIoN iN neutriNo OscillAtIon BecOme moRE defInite. IN particuLar if cp were conserved $\NU$ transition prOBaBILiTIes wOulD be the same aS $\bar{\NU}$ traNsitIOn ProBAbiliTies wHiLE tHE occurence therein of cP ViolatIon maKes these diffeRent [@Kayser]. tHIS differeNcc iS DePEndent on the neuTrino OscillatioN ParameteR $L/E$, $L$ bEing the dIstance of TRAvel from CreAtiOn tO deTECtIon and $E$ the eneRGY of tHe Initial NeuTrino: anD it Is aLso SenSiTive to the Value of tHe SmAlL rAtiO of neUTrino masS sQuaReD diFfereNCes. TheRe resUlts A cOmPLicAted 2-varIAbLE DepeNdEnCe in AddItIon to The lINeaR dependEnce on the LepTOnic jaRlSkog parAmeter, $J_{lep}$[@HSW]. paTtern for thE dIffErence BETween $\nu$ aNd $\bar\nu$ oscillations =================================================================== The iNPut to thE foRmula For nEutrino trAnsItion pRobABilitiEs is laRgely FrOm tHE MixinG MAtRix ElEments $U_{\alpHA I}$ whEre $\alPhA$ is oNe of the 3 Flavour indices and $i$ ONe oF the 3 mass eigenStaTe inDICeS. FrOM tHEse 9 ElEMenTS Plaquettes [@BD] can Be construcTeD, ThEse being phASe iNvAriant pRoducts Of 2 $U$ elEMents muLtiplied bY products Of 2 $u^{\staR

I. A. Lubashevsky, A. A. Zemlyanov , Con tin uum d escr ipti on of anomalou s dif fusion on a comb struc ture, J . Exp . T heor. Phys.* *8 7 * * ( 19 98 ) 7 00 – 71 3. C . W . Gardi ner, Handb ook o f Stochastic Me thods: for Ph ysics, Chemi str y andth e N a tural Sc ience s, Spr i nger-V erlag, Be rl i n, 200 4 , 3-rde d it ion. --- abstract: ' A v i rtual graphica l cons tr u ct i o n i s m ade to sho wthe d i fferenc e b e t w een neutrino andanti-neutri n o o scilla ti ons in the pres en c e o f CP violat ionwith CPTconser v ation.' author: - | R.G. M o or ho use $^ b $\ $^ b $Un iversity o fGlasg ow,G l a s g ow G 128QQ, U.K. \ title: Note on a P a tte rn fr om CP Vio la tionin Neu trino O scillations --- In troductio n = == === == ====Thereisint erest i n the p o ssi bi l i t ythat CPT violation m a y o ccur and thens ho wi n neutri no os cill a t ion e xper i me nts[@BL] ,[@CDK ] ,[@ MINOS]. .Howeve rthi s m ay be , CPviolat ion is l ong e s tablished andi t is of impor t an c e t o see k i t in neutri no o s cill atio n r esu l ts. I f one t a ke s a conservative poi nt of vi ew th at nature con serves CPT , a nd there are 3g enerations ofneutr inos, then the cons equen ces of C P violati o n in neut rin o o sci lla t i on become mored e fini te . In pa rti cular i f C P w ere co ns erved $\n u$ trans it io npr oba bilit i es would b e t he sa me as $\bar{ \nu}$ tra ns it i onprobabi l it i e s wh il etheocc ur encether e inof CP v iolationmak e s th es ediffere nt [@Kayser]. T his differ en ccis dep e n dent onthe neutrino oscillatio n parame ter $L/E $, $ L$ beingthe dista nce of tra vel fr om cr ea tio n to de t e ct ion a nd $E$ the e ner gy of t he i nitialneutrino: and it i s al so sensitivetothev a lu e o f t h e s ma l l r a t io of neutrinomass squar ed di fferences. The re result s a com plica t ed 2-va riable de pendencein add i t ion to the li near dep endence o n thel ep tonic Ja rlskog p ara meter , $J_{ l ep} $[@HS W]. P at tern f or th edifferen ce between $\nu$ and $\ bar\nu $ osc ill ations == === = === ========= ==== ========== === === ===== === = ===== ==== = == === = ==== The input tot he fo r m ul a for neutr i n o tr ansit ion probab ilit ies is largely fr o m the mixing m atri x ele men t s $U _{ \alpha i}$ whe re$\ a l pha$ ison e of the 3flavourin d icesand $i $ oneof the3 ma s s eige nsta teindices.Fro mt hese 9el em e nts pl aque tt es [@B D] can be c o n structed, thesebeing p hasei nva riant p roducts of 2 $U$ eleme nts multipl ied by pro ducts of 2 $ U^ {\star

I. A._Lubashevsky, A._A. Zemlyanov, Continuum description_of anomalous_diffusion_on a_comb_structure, J. Exp._Theor. Phys. **87**_(1998) 700–713. C. W. Gardiner,_Handbook of Stochastic_Methods:_for Physics, Chemistry and the Natural Sciences, Springer-Verlag, Berlin, 2004, 3-rd edition. --- abstract: 'A_virtual_graphical construction_is_made_to show the difference between_neutrino and anti-neutrino oscillations in_the presence_of CP violation with CPT conservation.' author: - | __ R. G. Moorhouse$^b$\ _ $^b$University of Glasgow,Glasgow G12 8QQ, U.K.\ title: Note_on a Pattern from CP Violation_in Neutrino Oscillations --- Introduction ============ There_is_interest_in the possibility that_CPT violation may occur and then_show in neutrino oscillation experiments[@BL],[@CDK], [@MINOS].._However this may be, CP violation is_long established and it is of_importance to seek it in_neutrino oscillation_results. If one takes a_conservative point of_view that_nature conserves CPT,_and there are 3 generations of_neutrinos, then the_consequences of CP violation in neutrino_oscillation_become more definite._In_particular_if CP_were conserved $\nu$_transition_probabilities would_be_the same as $\bar{\nu}$ transition probabilities_while_the occurence therein of CP violation makes_these different [@Kayser]. This_differencc_is dependent on the_neutrino oscillation parameter $L/E$, $L$_being the distance of travel from_creation to_detection and_$E$ the energy of the initial neutrino: and it is also_sensitive to the value of the_small ratio of neutrino_mass squared_differences._There results a_complicated_2-variable dependence_in addition to the linear dependence on_the leptonic_Jarlskog parameter, $J_{lep}$[@HSW]. Pattern for the difference_between $\nu$ and $\bar\nu$_oscillations =================================================================== The_input to the formula for neutrino_transition probabilities is largely from the_mixing matrix elements $U_{\alpha i}$_where_$\alpha$_is one of the 3_flavour indices and $i$ one of_the 3 mass_eigenstate indices. From these 9 elements plaquettes_[@BD]_can be constructed, these being phase_invariant_products of 2 $U$ elements multiplied_by_products_of 2 $U^{\star

. Such a behavior is generally observed in superspin-glass or cluster-glass where the microscopic flipping time of the fluctuating magnetic entities (superspins) is much longer than that of atomic spin-glass, and therefore, the observation time in units of microscopic flipping time is relatively shorter in comparison of atomic spin-glass.[@Markovich; @Markovich1; @Jonsson; @Sankar] Because of this, the length scales probed during the experimental time scale are shorter and the condition for rejuvenation, i.e. the length scales probed during the experimental time scale are larger than the so called overlap length, is not satisfied.[@Jonsson] The absence of complete rejuvenation in ZFC memory may also occur if the probing field is strong enough to perturb the intrinsic non-equilibrium dynamics of the spin-glass system. Temperature-field cycling and relaxation ---------------------------------------- The relaxation experiments discussed in previous section are complemented with an intermediate negative thermal cycling with and without field change. Figure. \[fig:AgingRej\] (a) shows the result of temperature cycling without field change under ZFC protocol. Here the sample is cooled to 50 K in zero field, then 100 Oe field is applied and the magnetization is recorded as a function of time for $t_1$ s. Then keeping the field constant, the temperature is changed to 40 K and magnetization is recorded for $t_2$ s. Thereafter temperature is again brought to 50 K (without changing the field) and magnetization is taken for $t_3$ s. According to Sun et al., for superspin or cluster -glass, the relaxation during $t_3$ should be the continuation of the relaxation at $t_1$.[@Sun; @Bandyopadhyay; @Vijay] The inset of Fig. \[fig:AgingRej\] (a) shows the relaxation at $t_1$ and $t_3$. The relaxation during $t_3$ has the same functional form as during $t_1$ but is shifted upward. Figure. \[fig:AgingRej\] (b) exhibits the relaxation under similar protocol but with a zero field between $t_1$ and $t_3$. As clear from the inset of Fig. \[fig:AgingRej\] (b), the relaxation at $t_3$ is the continuation

. Such a behavior is generally observed in superspin - glass or bunch - looking glass where the microscopic flipping time of the fluctuate charismatic entities (superspins) is much longer than that of nuclear tailspin - glass, and therefore, the observation time in units of microscopic flipping prison term is relatively shorter in comparison of atomic tailspin - glass.[@Markovich; @Markovich1; @Jonsson; @Sankar ] Because of this, the length scales probed during the experimental time scale are shorter and the condition for rejuvenation, i.e. the duration scales probed during the experimental time plate are larger than the so call lap length, is not satisfied.[@Jonsson ] The absence of complete rejuvenation in ZFC memory may also occur if the probing field is potent enough to perturb the intrinsic non - equilibrium dynamics of the spin - glass system. Temperature - field cycling and liberalization ---------------------------------------- The relaxation experiments discussed in previous section are complemented with an intermediate negative thermal cycling with and without field change. human body.   \[fig: AgingRej\ ]   (a) shows the resultant role of temperature cycling without field change under ZFC protocol. Here the sample is cooled to 50   potassium in zero battlefield, then 100   Oe field is applied and the magnetization is commemorate as a function of time for $ t_1 $   s. Then keeping the field ceaseless, the temperature is changed to 40   K and magnetization is recorded for $ t_2 $   s. Thereafter temperature is again brought to 50   K (without change the plain) and magnetization is taken for $ t_3 $   s. According to Sun et al. , for superspin or cluster -glass, the relaxation during $ t_3 $ should be the continuation of the liberalization at $ t_1$.[@Sun; @Bandyopadhyay; @Vijay ] The inset of Fig.   \[fig: AgingRej\ ] (a) shows the relaxation at $ t_1 $ and $ t_3$. The liberalization during $ t_3 $ has the same functional form as during $ t_1 $ but is shift upward. design.   \[fig: AgingRej\ ]   (b) exhibits the easiness under similar protocol but with a zero battlefield between $ t_1 $ and $ t_3$. As clean from the inset of Fig.   \[fig: AgingRej\ ]   (b), the liberalization at $ t_3 $ is the continuation

. Sufh a behavior is generaluy observed in superspmn-glass or clusger-glass where the microscopmc foippibg time of the fluctuaging magnvtic entiries (wuperspins) is much longed thau vhat of atomic xpin-glass, dnd therefore, dhd lbservation time in units of microssopic fkioping time is telatpvqly ahorter in comparison of atomic spjn-glass.[@Karkovich; @Marlovich1; @Jonsson; @Sankar] Becakse lf this, the length scales proved qyring the exoerimental time scale zre shorter and the condition fur reluvenation, u.e. thf length scaoes pgobed during bne exparimentsl time scale sre laeger than the so callxd overlap length, is not sativfned.[@Jonsson] The absencw if cokplede rdhuvdnauioi ih ZFC lemkry may alao occur if the probing field os wtrong enough to pewttrb the intrinsic non-equilibrium dynamibs or the spin-glass system. Twmperature-field cyclijg and rejaxation ---------------------------------------- The relaxation experiments discussed in psevions seeblon xee complemented with an intermediate negative trsrkak cycling with and withouy virjd change. Figore. \[fig:AyjnfRej\] (a) shows the redult of tempwrature crclimg without field change undwr ZFC protobol. Yere the sample is cooled to 50 Y in zero field, then 100 Oe field is appmied and thf magnetiaxtion is recordea ax d function of time for $t_1$ s. Ehen keepmng tke field conxtant, ehe temperwture is changed to 40 K anf maguetizdtion is rfcorded for $t_2$ s. Thereafter tempeceture is agaim trolght to 50 K (withput changing ehe field) and kagnetivation is taken ror $t_3$ s. Eccording to Sun et al., fos superspin oc cluster -glaws, tye relabxtion during $t_3$ should bv uhe continyation of the relawatiov at $t_1$.[@Sun; @Bandyo'cdyyay; @Vijay] The onsdt jf Fmg. \[fig:WcingRej\] (a) shmws ghe telaxxtion au $t_1$ xnd $y_3$. The relaxation durhng $f_3$ has the same funvtlonal forn as duryng $t_1$ but is xhifted upward. Figkre. \[fij:AgingCej\] (b) echifits the relaxation under simimar protofol but with a sero field betwezn $t_1$ and $t_3$. As clear from the inset of Fij. \[fig:AgingRej\] (b), the relaxqtion at $t_3$ is the cpktinuation

. Such a behavior is generally observed or where the flipping time of is longer than that atomic spin-glass, and the observation time in units of flipping time is relatively shorter in comparison of atomic spin-glass.[@Markovich; @Markovich1; @Jonsson; @Sankar] of this, the length scales probed during the experimental time scale are shorter the for i.e. length scales probed during the experimental time scale are larger than the so called overlap length, not satisfied.[@Jonsson] The absence of complete rejuvenation in memory may also occur the probing field is strong to the intrinsic dynamics the system. Temperature-field cycling relaxation ---------------------------------------- The relaxation experiments discussed in previous section are complemented with an intermediate negative thermal cycling and without Figure. \[fig:AgingRej\] shows result temperature cycling without under ZFC protocol. Here the sample 50 K in zero field, then 100 Oe is applied the magnetization is recorded as a of time for $t_1$ s. Then keeping the constant, the temperature is changed to 40 K and magnetization is recorded for $t_2$ s. is again brought to K (without changing field) magnetization taken $t_3$ s. to Sun et al., for superspin or cluster -glass, the relaxation $t_3$ should be the continuation of the relaxation at $t_1$.[@Sun; The of Fig. \[fig:AgingRej\] shows the relaxation at and The relaxation during $t_3$ same form but shifted Figure. \[fig:AgingRej\] (b) exhibits relaxation under similar protocol but a zero field between from the inset of Fig. \[fig:AgingRej\] (b), the at $t_3$ is the continuation

. Such a behavior is generally oBserved in sUpersPin-GlaSs Or clUsteR-glass where the MIcroScopic flipping time of thE flucTuATing MAgNetic EntitieS (SuPERspInS) iS muCh LOnGer thAn tHat of atOmic spin-glAss, AnD therefore, thE ObServation tIme In units of micRosCopic fLiPpiNG time Is rElatiVely shORter in ComparisoN oF Atomic SPin-glasS.[@mArKoviCh; @Markovich1; @JonssoN; @saNKar] Because of thIs, the lEnGTh SCAleS prObed during ThE expeRImental TImE SCAle ARe shorter and tHe condition FOr rEjuvenAtIon, I.E. the leNgth sCaLEs pRobed during The eXperimentAl time SCale are LArger thAn the sO caLleD oveRLaP lEngTh, IS noT SaTisFIed.[@jonsson] THe AbSence Of coMPLETe reJuvEnatIon in zFC memory may aLso OccuR If tHe proBing fIeld Is StronG enougH to peRtUrb the intrinsic Non-eQuilibriuM dyNaMicS oF the sPIn-glasS sySteM. TemperAture-fiELd cYcLING aNd relaxation ---------------------------------------- The relAxATIoN experimEnts diSCuSsED in previOuS seCtioN ARe comPlemENtEd with an IntermEDiAtE negatiVe ThermaL cYclIng With aND witHout fiEld changE. FiguRE. \[fig:AgingRej\] (a) sHOws the result oF TeMPErATure CycLing without FielD ChanGe unDEr zFC PRotocOl. HerE tHE sAMple is cooled to 50 K in zeRo Field, tHen 100 Oe Field is applieD and the magNETIzation iS recORdED as a function of Time fOr $t_1$ s. Then keEPing the fIeld cOnstant, tHe temperaTURe is chanGed To 40 K And MagNETiZation is recorDED for $T_2$ s. thereafTer TemperaTurE is AgaIn bRoUght to 50 K (wiThout chaNgInG tHe FieLd) and MAgnetizaTiOn iS tAkeN for $t_3$ S. accordIng to sun eT aL., fOR suPerspin OR cLUSter -GlAsS, the RelAxAtion DuriNG $t_3$ sHould be The continUatIOn of ThE rElaxatiOn at $t_1$.[@Sun; @BandyOpAdhyay; @VijaY] THe iNset of fIG. \[fig:AginGRej\] (a) shows the relaxation AT $t_1$ and $t_3$. THe rElaxaTion During $t_3$ haS thE same fUncTIonal fOrm as dUring $T_1$ bUt iS SHifteD UPwArd. fiGure. \[fig:AgiNGrej\] (B) exhiBiTs thE relaxaTion under similar prOTocOl but with a zerO fiEld bETWeEn $t_1$ ANd $T_3$. as cLeAR frOM The inset of Fig. \[fiG:AgingRej\] (b), ThE ReLaxation at $T_3$ Is tHe ContinuAtion

. Such a behavior is gener ally obser ved i n s upe rs pin- glas s or cluster-g l asswhere the microscopicflipp in g tim e o f the fluctu a ti n g ma gn et icen t it ies ( sup erspins ) is muchlon ge r than thato fatomic spi n-g lass, and th ere fore,th e o b serva tio n tim e in u n its of microsco pi c flipp i ng time i srela tively shorter in co m parison of ato mic sp in - gl a s s.[ @Ma rkovich; @ Ma rkovi c h1; @Jo n ss o n ; @S a nkar] Because of this, t h e l engthsc ale s probe d dur in g th e experimen taltime scal e ares hortera nd thecondit ion fo r re j uv en ati on , i. e .the len gth scal es p robed dur i n g theexp erim ental time scale a relarg e r t han t he so cal le d ove rlap l ength ,is not satisfie d.[@ Jonsson]The a bse nc e ofc omplet e r eju venatio n in ZF C me mo r y ma y also occur if th ep r ob ing fiel d is s t ro ng enough t oper turb t he in trin s ic non-equ ilibri u mdy namicsof the s pi n-g las s sys t em. Tempe rature-f ieldc ycling and rel a xation ------ - -- - - -- - ---- --- ----------- ---- - --- The re lax a tionexper im e nt s discussed in previ ou s sect ion a re complement ed with an i n termedia te n e ga t ive thermal cy cling with andw ithout f ieldchange.Figure. \ [ f ig:Aging Rej \](a) sh o w sthe result of t empe ra ture cy cli ng with out fi eld ch an ge underZFC prot oc ol .He rethe s a mple isco ole dto50 Ki n zero fiel d, t he n1 00Oe fiel d i s appl ie dandthe m agnet izat i onis reco rded as a fu n ctio nof time f or $t_1$ s. T he n keepingth e f ield c o n stant, t he temperature is chang e d to 40  Kand m agne tizationisrecord edf or $t_ 2$ s.There af ter t emper a t ur e i sagain brou g h t t o 50K(wit hout ch anging the field)a ndmagnetization is tak e n f or$ t_ 3 $ s .A cco r d ing to Sun et a l., for su pe r sp in or clus t er-g lass, t he rela xatio n during $t_3$ sh ould be t he con t i nua tion of th e relaxa tion at $ t _1$.[ @ Su n; @B and yopadh ya y;@Vija y] The ins et of Fig.\[ fig:Ag ingRe j\ ] (a) sh ows the relaxation at $ t_1$ a nd $t _3$ . The rel axa t ion during $ t_3$ has the s ame fu nctio nal formas d u ri ng$ t_1$buti s shifted up war d . F igure. \[fi g : A gin gRej\ ] ( b ) exhi bits the relaxation u n der similar pr otoc o l bu t w i th a z ero field betw een $ t _ 1$ and $ t_ 3$. As clea r from t he inset of Fi g. \[f ig:Agin g R ej \ ] (b), the re laxationat$t _ 3$ is t he c o ntinua tion

. Such_a behavior_is generally observed in_superspin-glass or_cluster-glass_where the_microscopic_flipping time of_the fluctuating magnetic_entities (superspins) is much_longer than that_of_atomic spin-glass, and therefore, the observation time in units of microscopic flipping time is_relatively_shorter in_comparison_of_atomic spin-glass.[@Markovich; @Markovich1; @Jonsson; @Sankar]_Because of this, the length_scales probed_during the experimental time scale are shorter and_the_condition for rejuvenation,_i.e. the length scales probed during the experimental time_scale are larger than the so_called overlap length,_is_not_satisfied.[@Jonsson] The absence of_complete rejuvenation in ZFC memory may_also occur if the probing field_is strong enough to perturb the intrinsic_non-equilibrium dynamics of the spin-glass system. Temperature-field_cycling and relaxation ---------------------------------------- The relaxation experiments_discussed in_previous section are complemented with_an intermediate negative_thermal cycling_with and without_field change. Figure. \[fig:AgingRej\] (a) shows the result_of temperature cycling_without field change under ZFC protocol._Here_the sample is_cooled_to_50 K in_zero field, then_100 Oe_field is_applied_and the magnetization is recorded as_a_function of time for $t_1$ s. Then keeping_the field constant, the_temperature_is changed to 40 K_and magnetization is recorded for_$t_2$ s. Thereafter temperature is again brought_to 50 K_(without changing_the field) and magnetization is taken for $t_3$ s. According to Sun_et al., for superspin or cluster_-glass, the relaxation during_$t_3$ should_be_the continuation of_the_relaxation at_$t_1$.[@Sun; @Bandyopadhyay; @Vijay] The inset of Fig. \[fig:AgingRej\]_(a) shows_the relaxation at $t_1$ and $t_3$._The relaxation during $t_3$_has_the same functional form as during_$t_1$ but is shifted upward. Figure. \[fig:AgingRej\] (b)_exhibits the relaxation under similar_protocol_but_with a zero field between_$t_1$ and $t_3$. As clear from_the inset of_Fig. \[fig:AgingRej\] (b), the relaxation at $t_3$ is the_continuation

$V$, and an injective homomorphism $f:\A\to \wp(V)$ such that $\bigcap_{x\in X_i} f(x)=\emptyset$ for each $i\in \lambda$. The idea is exactly like the previous proof, except that now we have a union of $<{}^{\omega}2$ no where dense sets; the required ultrafilter to build the representation we need lies outside this union. $MA$ offers solace here because it implies that such a union can be written as a countable union, and again the Baire category theorem readily applies. But without $MA$, if the given algebra is countable and completely additive, and if the types are maximal, then we can also omit $<{} ^{\omega}2$ types. This [*is indeed* ]{} provable in $ZFC$, without any additional assumptions at all. For brevity, when we have an omitting types theore, like the previos one, we say that the hypothesis implies existence of a representation omitting the given set of non-principl types. Let $\A\in SA_n$ be a countable, let $\lambda<{}^{\omega}2$ and $\bold F=(F_i: i<\lambda)$ be a family of non principal ultrafilters. Then there is a single representation that omits these non-principal ultrafilters. One can easily construct two representations that overlap only on principal ultrafilters [@h]. With a pinch of diagonalization this can be pushed to countable many. But even more, using ideas of Shelah [@Shelah] thm 5.16, this this can be further pushed to $^{\omega}2$ many representations. Such representations are not necessarily pair-wise distinct, but they can be indexed by a set $I$ such that $|I|={}^{\omega}2$, and if $i, j\in I$, are distinct and there is an ultrafilter that is realized in the representations indexed by $i$ and $j$, then it is principal. Note that they can be the same representation. Now assume for contradiction that there is no representation omitting the given non-principal ultrafilters. Then for all $i<{}^{\omega}2$, there exists $F$ such that $F$ is realized in $\B_i$. Let $\psi:{}^{\omega}2\to \wp(\bold F)$, be defined by $\psi(i)=\{

$ V$, and an injective homomorphism $ f:\A\to \wp(V)$ such that $ \bigcap_{x\in X_i } f(x)=\emptyset$ for each $ i\in \lambda$. The idea is exactly like the previous proof, except that nowadays we have a coupling of $ < { } ^{\omega}2 $ no where dense sets; the required ultrafilter to build the theatrical performance we need lies outside this union. $ MA$ offer solace here because it implies that such a coupling can be written as a countable union, and again the Baire class theorem readily applies. But without $ MA$, if the given algebra is countable and completely linear, and if the types are maximal, then we can also exclude $ < { } ^{\omega}2 $ types. This [ * is indeed * ] { } provable in $ ZFC$, without any additional premise at all. For brevity, when we have an omitting types theore, like the previos one, we read that the hypothesis implies existence of a representation exclude the given set of non - principl types. Let $ \A\in SA_n$ be a countable, let $ \lambda<{}^{\omega}2 $ and $ \bold F=(F_i: i<\lambda)$ be a syndicate of non principal ultrafilters. Then there is a single representation that omits these non - principal ultrafilters. One can easily construct two representations that overlap only on principal ultrafilters [ @h ]. With a pinch of diagonalization this can be pushed to countable many. But even more, using ideas of Shelah [ @Shelah ] thm 5.16, this this can be further pushed to $ ^{\omega}2 $ many representation. Such representations are not necessarily couple - fresh distinct, but they can be index by a set $ I$ such that $ |I|={}^{\omega}2 $, and if $ i, j\in I$, are distinct and there is an ultrafilter that is realized in the representations indexed by $ i$ and $ j$, then it is principal. notice that they can be the same representation. Now assume for contradiction that there is no representation omitting the given non - principal ultrafilters. Then for all $ i<{}^{\omega}2 $, there exists $ F$ such that $ F$ is realized in $ \B_i$. Let $ \psi:{}^{\omega}2\to \wp(\bold F)$, be define by $ \psi(i)=\ {

$V$, wnd an injective homomorkhism $f:\A\to \wp(V)$ suey that $\bigcal_{x\in X_i} w(x)=\emptyset$ for each $i\in \lambva$. Thw ideq is exactly like the orevious iroof, excwpt uhat now we have e union of $<{}^{\omefw}2$ no xhere dense setx; the requhred ultrafiltar tl build the representation we need jies ouyslde this union. $MA$ psfera solace here because it implies tgat suci a union can br written as a countable ujion, and again the Baige category thejeem readily xpplies. But without $MA$, jf the given algebra is countabue anb completeli zdfhtive, and ih the nypes are maxlkal, than we csn also omit $<{} ^{\pmeja}2$ ttpes. This [*is indeed* ]{} 'rovable in $ZFC$, withjut any agdntional assumptions ar qll. Fmr bsevigt, wfen wx hzve an ommtting typea theore, lije the previos one, ee way that the gypothqsys implies existence of a representatiot ojitting the given set od non-principl types. Lej $\A\in SA_n$ fe a countable, let $\lambda<{}^{\omega}2$ and $\bold F=(F_i: i<\lambga)$ be x fcnily ud jon principal ultrafilters. Then there is a sigflt rvpresentation thab omits these non-ptijcokal ultrafiltets. One cch sasily construct tao reprgsentarions thau ovetlap only on principal ultrqfilters [@h]. Wpth q pinch of diagonapization thns can be pished to countable many. But even more, ksing idezr of Shelah [@Shelxh] nhm 5.16, this this can be further [ushed to $^{\omeya}2$ many feprgsentatyons. Such gepresentations are not jecesdasily pair-wlse distinct, but they can be invxxed by a set $I$ subh that $|I|={}^{\jmega}2$, and if $i, j\in I$, are distincj and thexe is xn ultrafimter thet is realizqd in the repsgsentations iidexed by $i$ abd $j$, then ig is principal. Mote that they can ve the same represcntatkkn. Now assume fir xontradiction tnat thqrv ix gm representadion omkyting the given kon-orinvipal ultrafilters. Than fkr all $i<{}^{\omega}2$, therr cxists $F$ wuch thae $F$ is realizrd in $\B_i$. Let $\psi:{}^{\omtga}2\to \xp(\bold F)$, be desined by $\psi(i)=\{

$V$, and an injective homomorphism $f:\A\to \wp(V)$ $\bigcap_{x\in f(x)=\emptyset$ for $i\in \lambda$. The previous except that now have a union $<{}^{\omega}2$ no where dense sets; the ultrafilter to build the representation we need lies outside this union. $MA$ offers here because it implies that such a union can be written as a union, again Baire theorem readily applies. But without $MA$, if the given algebra is countable and completely additive, and the types are maximal, then we can also $<{} ^{\omega}2$ types. This indeed* ]{} provable in $ZFC$, any assumptions at For when have an omitting theore, like the previos one, we say that the hypothesis implies existence of a representation omitting the set of Let $\A\in be countable, $\lambda<{}^{\omega}2$ and $\bold be a family of non principal is a single representation that omits these non-principal One can construct two representations that overlap only principal ultrafilters [@h]. With a pinch of diagonalization can be pushed to countable many. But even more, using ideas of Shelah [@Shelah] thm this can be further to $^{\omega}2$ many Such are necessarily distinct, but can be indexed by a set $I$ such that $|I|={}^{\omega}2$, and $i, j\in I$, are distinct and there is an ultrafilter realized the representations indexed $i$ and $j$, then is Note that they can same Now that is representation omitting the given ultrafilters. Then for all $i<{}^{\omega}2$, exists $F$ such that Let $\psi:{}^{\omega}2\to \wp(\bold F)$, be defined by $\psi(i)=\{

$V$, and an injective homomorphiSm $f:\A\to \wp(V)$ sUch thAt $\bIgcAp_{X\in X_I} f(x)=\eMptyset$ for each $I\In \laMbda$. The idea is exactly liKe the PrEViouS PrOof, exCept thaT NoW WE haVe A uNioN oF $<{}^{\OmEga}2$ no WheRe dense Sets; the reqUirEd Ultrafilter tO BuIld the reprEseNtation we neeD liEs outsIdE thIS unioN. $MA$ OfferS solacE Here beCause it imPlIEs that SUch a uniON CaN be wRitten as a countablE UnIOn, and again the BAire caTeGOrY THeoRem Readily appLiEs. But WIthout $Ma$, If THE GivEN algebra is couNtable and coMPleTely adDiTivE, And if tHe typEs ARe mAximal, then wE can Also omit $<{} ^{\oMega}2$ tyPEs. This [*iS Indeed* ]{} pRovablE in $zFC$, WithOUt AnY adDiTIonAL aSsuMPtiOns at all. foR bRevitY, wheN WE HAve aN omIttiNg typEs theore, like tHe pReviOS onE, we saY that The hYpOthesIs implIes exIsTence of a represeNtatIon omittiNg tHe GivEn Set of NOn-prinCipL tyPes. Let $\A\In SA_n$ be A CouNtABLE, lEt $\lambda<{}^{\omega}2$ and $\boLd f=(f_I: i<\Lambda)$ be A familY Of NoN PrincipaL uLtrAfilTERs. TheN theRE iS a single RepresENtAtIon that OmIts theSe Non-PriNcipaL UltrAfilteRs. One can EasilY Construct two rePResentations tHAt OVErLAp onLy oN principal uLtraFIlteRs [@h]. WITh A piNCh of dIagonAlIZaTIon this can be pushed tO cOuntabLe manY. But even more, uSing ideas oF sHElah [@ShelAh] thM 5.16, ThIS this can be furtHer puShed to $^{\omegA}2$ Many reprEsentAtions. SuCh represeNTAtions arE noT neCesSarILY pAir-wise distinCT, But tHeY can be iNdeXed by a sEt $I$ SucH thAt $|I|={}^{\OmEga}2$, and if $i, J\in I$, are dIsTiNcT aNd tHere iS An ultrafIlTer ThAt iS realIZed in tHe repReseNtAtIOns Indexed BY $i$ AND $j$, thEn It Is prIncIpAl. NotE thaT TheY can be tHe same repResENtatIoN. NOw assumE for contradicTiOn that therE iS no RepresENTation omItting the given non-princiPAl ultraFilTers. THen fOr all $i<{}^{\omeGa}2$, tHere exIstS $f$ such tHat $F$ is RealiZeD in $\b_I$. let $\psI:{}^{\OMeGa}2\tO \wP(\bold F)$, be deFINed By $\psi(I)=\{

$V$, and an injective hom omorphism$f:\A \to \w p( V)$such that $\bigcap _ {x\i n X_i} f(x)=\emptyset$ forea c h $i \ in \lam bda$. T he i dea i sexa ct l ylikethe previo us proof,exc ep t that now w e h ave a unio n o f $<{}^{\ome ga} 2$ nowh ere dense se ts; t he req u ired u ltrafilte rt o buil d the re p r es enta tion we need lies ou t side this unio n. $MA $o ff e r s s ola ce here be ca use i t implie s t h a t su c h a union can be written asa coun ta ble union, andag a inthe Baire c ateg ory theor em rea d ily app l ies. B ut wit hou t $ MA$, if t hegi v ena lg ebr a is countab le a nd co mple t e l y add iti ve,and i f the types a remaxi m al, then we c an a ls o omi t $<{} ^{\o me ga}2$ types. Th is [ *is indee d*]{ } p ro vable in $ZF C$, wi thout a ny addi t ion al a s su mptions at all. Fo rb r ev ity, whe n we h a ve a n omittin gtyp es t h e ore,like th e previo s one, we s ay that t he hyp ot hes isimpli e s ex istenc e of a r epres e ntation omitti n g the given s e to f n o n-pr inc ipl types. Let $\A\ in S A _n $ b e a co untab le , l e t $\lambda<{}^{\ome ga }2$ an d $\b old F=(F_i: i <\lambda)$ b e a famil y of no n principal ult rafil ters. Then there is a si ngle rep resentati o n that om its th ese no n - pr incipal ultra f i lter s. One c aneasilycon str uct tw orepresent ations t ha tov er lap only on princ ip alul tra filte r s [@h] . Wit h api nc h of diagon a li z a tion t hi s ca n b epushe d to cou ntablemany. But ev e n mo re ,using i deas of Shela h[@Shelah]th m 5 .16, t h i s this c an be further pushed to $^{\ome ga} 2$ ma ny r epresenta tio ns. Su chr eprese ntatio ns ar enot n ecess a r il y p ai r-wise dis t i nct , but t heycan beindexed by a set $ I $ s uch that $|I| ={} ^{\o m e ga }2$ , a n d i f$ i,j \ in I$, are dist inct and t he r eis an ultr a fil te r thatis real izedi n the r epresenta tions ind ex ed b y $i$ and $j$,then itis princi p al. N o te that th ey can b e t he sa me rep r ese ntati on. No wassume forco ntradict ion that there is no re presen tatio n o mitting t heg ive n non-pri ncip al ultrafi lte rs. The n f o r all $i< { }^ {\o m ega}2 $, t h ere exist s $ F$s u ch that $F$ i s r eal izedin$ \B_i$. Let $\psi:{}^{\omega } 2\to \wp(\bold F)$ , bedef i nedby $\psi(i)=\{

$V$,_and an_injective homomorphism $f:\A\to \wp(V)$_such that_$\bigcap_{x\in_X_i} f(x)=\emptyset$_for_each $i\in \lambda$. The_idea is exactly_like the previous proof,_except that now_we_have a union of $<{}^{\omega}2$ no where dense sets; the required ultrafilter to build_the_representation we_need_lies_outside this union. $MA$ offers_solace here because it implies_that such_a union can be written as a countable_union,_and again the_Baire category theorem readily applies. But without $MA$, if the_given algebra is countable and completely_additive, and if_the_types_are maximal, then we_can also omit $<{} ^{\omega}2$ types._This [*is indeed* ]{} provable in_$ZFC$, without any additional assumptions at all._For brevity, when we have an_omitting types theore, like the_previos one,_we say that the hypothesis_implies existence of_a representation_omitting the given_set of non-principl types. Let $\A\in SA_n$_be a countable,_let $\lambda<{}^{\omega}2$ and $\bold F=(F_i: i<\lambda)$_be_a family of_non_principal_ultrafilters. Then_there is a_single_representation that_omits_these non-principal ultrafilters. One can easily construct_two_representations that overlap only on principal ultrafilters_[@h]. With a pinch_of_diagonalization this can be_pushed to countable many. But_even more, using ideas of Shelah_[@Shelah] thm_5.16, this_this can be further pushed to $^{\omega}2$ many representations. Such representations_are not necessarily pair-wise distinct, but_they can be indexed_by a_set_$I$ such that_$|I|={}^{\omega}2$,_and if_$i, j\in I$, are distinct and there_is an_ultrafilter that is realized in the_representations indexed by $i$_and_$j$, then it is principal. Note_that they can be the same_representation. Now assume for contradiction_that_there_is no representation omitting the_given non-principal ultrafilters. Then for all $i<{}^{\omega}2$,_there exists $F$_such that $F$ is realized in $\B_i$._Let_$\psi:{}^{\omega}2\to \wp(\bold F)$, be defined by_$\psi(i)=\{

rightarrow \Delta_0 $. As $ \tilde{ \mu } $ increases, $ \tilde{ \Delta }_0 $ reaches a maximum at an optimum value of the chemical potential and $ \tilde{ \Delta }_0 $ decreases as $ \tilde{ \mu } $ increases even further. For the case $ \tilde{ \lambda }_{ \mbox{\scriptsize{sc}} } < 1 $, our results are shown in Figure \[FigD0vsmu\_LscLT1\_2\]. Starting at $ \tilde{ \mu } = 0 $, the system is in the normal state. However, as $ \tilde{ \mu } $ increases, the system asymptotically becomes superconducting and the order parameter also increases up to a maximum value at an optimal chemical potential. As $ \tilde{ \mu } $ increases even further, $ \tilde{ \Delta }_0 $ decreases and the energy gap displays a dome-shaped plot. Our results are consistent with [@Fukushima2007], where chiral and diquark condensates are calculated for two-color and two-flavor QCD. For a certain choice of parameters, the authors obtain the same dispersion relation given by Eq. (\[Ej\]) and their numerical results for $ \Delta $ also display a dome-shaped plot, as can be seen in Fig. 1 of [@Fukushima2007], for the choice of parameters I, referred as the [*weak-coupling case*]{}. Since the energy gap and the superconducting critical temperature $ T_{ c }$ are proportional, our results qualitatively reproduce the superconducting phase diagram of several compounds, like 122 pnictides and cuprate superconductors, where the critical temperature displays a characteristic dome as charge carriers are doped into the system. A dome-like structure of the superconducting phase for strongly interacting two-dimensional Dirac fermions has been previously obtained in [@Smith2009], where the authors observed a dome-shaped plot of the superconducting phase at intermediate filling fractions, surrounded by the normal phase for fillings close to unity or zero, which is consistent to our results. We conclude this section with a final remark: contrary to what happens at $ \mu = 0 $, where superconductivity appears in the system only when $ \lambda_{ \mbox{\scriptsize{sc}} } > \lambda_{c } $, at finite chemical potential, we could always find a finite $ \tilde{ \mu } $ that provided nonzero superconducting gaps, even at small values of $ \lambda_{ \

rightarrow \Delta_0 $. As $ \tilde { \mu } $ increases, $ \tilde { \Delta } _ 0 $ reaches a maximum at an optimum value of the chemical potential and $ \tilde { \Delta } _ 0 $ decrease as $ \tilde { \mu } $ increase even further. For the case $ \tilde { \lambda } _ { \mbox{\scriptsize{sc } } } < 1 $, our results are prove in Figure   \[FigD0vsmu\_LscLT1\_2\ ]. Starting at $ \tilde { \mu } = 0 $, the system is in the normal state of matter. However, as $ \tilde { \mu } $ increases, the system asymptotically become superconducting and the order parameter besides increases up to a maximum value at an optimal chemical potential. As $ \tilde { \mu } $ increase even further, $ \tilde { \Delta } _ 0 $ decreases and the energy gap displays a attic - shaped plot. Our results are consistent with [ @Fukushima2007 ], where chiral and diquark condensation are calculated for two - color and two - spirit QCD. For a certain option of parameters, the generator obtain the same dispersion relation back given by Eq.   (\[Ej\ ]) and their numerical results for $ \Delta $ also display a dome - shaped plot, as can be seen in Fig. 1 of [ @Fukushima2007 ], for the choice of parameters I, referred as the [ * weak - coupling case * ] { }. Since the energy gap and the superconducting critical temperature $ T _ { c } $ are proportional, our consequence qualitatively reproduce the superconducting phase diagram of several compound, like 122 pnictides and cuprate superconductors, where the critical temperature displays a characteristic dome as charge carriers are dope into the system. A dome - like structure of the superconducting phase for powerfully interacting two - dimensional Dirac fermions has been previously obtained in [ @Smith2009 ], where the authors observed a dome - shaped plot of the superconducting phase at intermediate filling fractions, surrounded by the normal phase for fillings cheeseparing to unity or zero, which is reproducible to our results. We conclude this section with a final remark: contrary to what find at $ \mu = 0 $, where superconductivity appears in the system only when $ \lambda _ { \mbox{\scriptsize{sc } } } > \lambda_{c } $, at finite chemical potential, we could always find a finite $ \tilde { \mu } $ that provided nonzero superconducting gaps, even at little values of $ \lambda _ { \

rigjtarrow \Delta_0 $. As $ \tilde{ \mu } $ increases, $ \tilde{ \Delta }_0 $ reachds a maximum at an optimum velue of tye chemical potential xnd $ \tildv{ \Delta }_0 $ decceases as $ \tilde{ \mu } $ ingxeasea eveu hurther. For the gase $ \tilde{ \lambda }_{ \mbox{\swrkpcsize{sc}} } < 1 $, our results are shown in Figure \[GihD0vsmu\_LscLT1\_2\]. Stwrtimd at $ \tilde{ \mu } = 0 $, the system is in ths norman state. Howevrr, as $ \tilde{ \mu } $ increased, thf system asymptotifally becomga stperconductine and the order paramejer also increases up to a maximjm vakue at an iprimwn chemical 'otentpal. As $ \tilde{ \mu } $ inwreases even further, $ \tinde{ \Delta }_0 $ decreases anv the energy gap disklays a doke-ahaped plot. Our rewuots ate cotsisgwnt wiuh [@Hukhshima2007], whxre chiral znd diquark condensates are cakctoated for two-dolor wnq two-flavor QCD. For a certain choice of padameters, the authors obrain the same disperslon relatyon given by Eq. (\[Ej\]) and their numerical results for $ \Delva $ aosj eidplay a dome-shaped plot, as can be seen in Fig. 1 og [@Nukushima2007], for thc choice of parameyegs Y, referred as the [*wzzk-doupling case*]{}. Since the engrgy gqp and tht suprrconducting critical tempeeature $ T_{ c }$ qre proportional, obr results qbalitajively reproduce the superconbuctinf phase diahram of ssxeral compounds, uikv 122 ptictides and cuprate supersonductorw, whzre the zritocal tqmperature displays a characteristlc dole as charge carriers are doped into the system. A dome-like sdrubture of che suierconducting prase for stronyly intexactine two-dimenaional Virac fermiogs has been psgviously obtamned in [@Siith2009], wheee the xjthors observec a dome-skcped plot of the supercondugting lhase at intermzbiqte filling fravtiuns, slrrpugged by the nmrmau pfsse fur fillings zlosr to unity or zero, wvich is consistent to puv results. Qe concltde this sectoon with a final rtmark: rontracy to ehaj happens at $ \mu = 0 $, where superdonductivlty appears in ehe wystem only chen $ \lambda_{ \mbox{\scriptsize{sc}} } > \lambda_{c } $, at finite chemixal potential, we coold always finv a figite $ \tilge{ \mu } $ that providee nonzero supercokducting gaps, even at amall ealued of $ \lambda_{ \

rightarrow \Delta_0 $. As $ \tilde{ \mu increases, \tilde{ \Delta $ reaches a of chemical potential and \tilde{ \Delta }_0 decreases as $ \tilde{ \mu } increases even further. For the case $ \tilde{ \lambda }_{ \mbox{\scriptsize{sc}} } < $, our results are shown in Figure \[FigD0vsmu\_LscLT1\_2\]. Starting at $ \tilde{ \mu = $, system in the normal state. However, as $ \tilde{ \mu } $ increases, the system asymptotically becomes and the order parameter also increases up to maximum value at an chemical potential. As $ \tilde{ } increases even $ \Delta $ decreases and energy gap displays a dome-shaped plot. Our results are consistent with [@Fukushima2007], where chiral and diquark condensates calculated for two-flavor QCD. a choice parameters, the authors same dispersion relation given by Eq. numerical results for $ \Delta $ also display dome-shaped plot, can be seen in Fig. 1 [@Fukushima2007], for the choice of parameters I, referred the [*weak-coupling case*]{}. Since the energy gap and the superconducting critical temperature $ T_{ c proportional, our results qualitatively the superconducting phase of compounds, 122 and cuprate where the critical temperature displays a characteristic dome as charge carriers doped into the system. A dome-like structure of the superconducting strongly two-dimensional Dirac fermions been previously obtained in where authors observed a dome-shaped the phase fractions, by normal phase for fillings to unity or zero, which consistent to our results. a final remark: contrary to what happens at \mu = 0 $, where superconductivity appears the system only when $ \lambda_{ \mbox{\scriptsize{sc}} } > \lambda_{c } $, finite chemical could always find a finite $ \tilde{ \mu $ that provided nonzero gaps, even at small values of $ \lambda_{ \

rightarrow \Delta_0 $. As $ \tilde{ \mu } $ iNcreases, $ \tiLde{ \DeLta }_0 $ ReaChEs a mAximUm at an optimum vALue oF the chemical potential aNd $ \tilDe{ \dElta }_0 $ DEcReaseS as $ \tildE{ \Mu } $ INCreAsEs EveN fURtHer. FoR thE case $ \tiLde{ \lambda }_{ \mBox{\ScRiptsize{sc}} } < 1 $, ouR ReSults are shOwn In Figure \[FigD0VsmU\_LscLT1\_2\]. stArtINg at $ \tIldE{ \mu } = 0 $, thE systeM Is in thE normal stAtE. howeveR, As $ \tilde{ \MU } $ InCreaSes, the system asympTOtICally becomes suPerconDuCTiNG And The Order paramEtEr alsO IncreasES uP TO A maXImum value at an Optimal chemICal PotentIaL. As $ \TIlde{ \mu } $ IncreAsES evEn further, $ \tiLde{ \DElta }_0 $ decreAses anD The enerGY gap disPlays a DomE-shAped PLoT. OUr rEsULts ARe ConSIstEnt with [@FUkUsHima2007], wHere CHIRAl anD diQuarK condEnsates are calCulAted FOr tWo-colOr and Two-fLaVor QCd. For a cErtaiN cHoice of parameteRs, thE authors oBtaIn The SaMe disPErsion RelAtiOn given By Eq. (\[Ej\]) aND thEiR NUMeRical results for $ \DelTa $ ALSo Display a Dome-shAPeD pLOt, as can bE sEen In FiG. 1 OF [@FukuShimA2007], FoR the choiCe of paRAmEtErs I, refErRed as tHe [*WeaK-coUplinG Case*]{}. since tHe energy Gap anD The superconducTIng critical teMPeRATuRE $ T_{ c }$ aRe pRoportional, Our rESultS quaLItAtiVEly reProduCe THe SUperconducting phase DiAgram oF seveRal compounds, lIke 122 pnictidES ANd cupratE supERcONductors, where tHe criTical tempeRAture disPlays A charactEristic doME As charge CarRieRs aRe dOPEd Into the system. a DOme-lIkE structUre Of the suPerConDucTinG pHase for stRongly inTeRaCtInG twO-dimeNSional DiRaC feRmIonS has bEEn prevIouslY obtAiNeD In [@SMith2009], wheRE tHE AuthOrS oBserVed A dOme-shAped PLot Of the suPerconducTinG PhasE aT iNtermedIate filling frAcTions, surroUnDed By the nORMal phase For fillings close to unity OR zero, whIch Is conSistEnt to our rEsuLts. We cOncLUde thiS sectiOn witH a FinAL RemarK: COnTraRy To what happENS at $ \Mu = 0 $, wheRe SupeRconducTivity appears in the SYstEm only when $ \lamBda_{ \Mbox{\SCRiPtsIZe{SC}} } > \laMbDA_{c } $, aT FInite chemical poTential, we cOuLD aLways find a FIniTe $ \Tilde{ \mu } $ That proVided NOnzero sUperconduCting gaps, EvEn at SMAll Values of $ \laMbda_{ \

rightarrow \Delta_0 $. As$ \tilde{\mu } $inc re ases , $\tilde{ \Delta }_0$ reaches a maximum at an o pt i mumv al ue of the ch e mi c a l p ot en tia la nd $ \t ild e{ \Del ta }_0 $ d ecr ea ses as $ \ti l de { \mu } $inc reases evenfur ther. F ort he ca se$ \ti lde{ \ l ambda}_{ \mbox {\ s cripts i ze{sc}} } < 1 $ , our results are sh o wn in Figure \ [FigD0 vs m u\ _ L scL T1\ _2\]. Star ti ng at $ \tild e {\ m u }= 0 $, the sys tem is in t h e n ormalst ate . Howev er, a s$ \t ilde{ \mu } $ i ncreases, the s y stem as y mptotic ally b eco mes sup e rc on duc ti n g a n dthe ord er param et er also inc r e a s es u p t o amaxim um value at a n o ptim a l c hemic al po tent ia l. As $ \ti lde{\m u } $ increases eve n further , $ \ til de { \De l ta }_0 $dec reasesand the ene rg y g ap displays a dome-s ha p e dplot. O ur res u lt sa re consi st ent wit h [@Fuk ushi m a2 007], wh ere ch i ra land diq ua rk con de nsa tes arec alcu latedfor two- color and two-flavor QCD. For a ce r ta i n c h oice of parameters , th e aut hors ob tai n thesamedi s pe r sion relation given b y Eq.(\[Ej \]) and their numerical r e sults fo r $\ De l ta $ also disp lay a dome-shap e d plot,as ca n be see n in Fig. 1 of [@Fu kus him a20 07] , fo r the choiceo f par am eters I , r eferred as th e [ *we ak -coupling case*]{ }. Si nc e t he en e rgy gapan d t he su perco n ductin g cri tica lte m per ature $ T_ { c }$ a re pro por ti onal, our res ults qu alitative lyr epro du ce the su perconducting p hase diagr am of sever a l compoun ds, like 122 pnictidesa nd cupr ate supe rcon ductors,whe re the cr i ticaltemper ature d isp l a ys ac h ar act er istic dome a s c harge c arri ers are doped into the sy s tem . A dome-lik e s truc t u re of th e su pe r con d u cting phase for stronglyin t er acting two - dim en sionalDirac f ermio n s has b een previ ously obt ai nedi n [@ Smith2009] , wherethe autho r s obs e rv ed adom e-shap ed pl ot of the s u per condu ctingph ase at inte rm ediate f illing fractions, surro undedby th e n ormal pha sef orfillingsclos e to unity or ze ro, w hic h is c onsi s te ntt o our res u lts. Wec on clu d e t his section w i tha fin alr emark: con trary to what hap p ens at $ \mu = 0 $ , whe res uper co nductivity app ear si n the sys te m only when $ \lam bd a _{ \m box{\s cripts ize{sc} } }> \lamb da_{ c } $, at fi nit ec hemical p ot e ntial, weco uld al ways f i nd a f inite $ \tilde{\mu } $ that pro vided n onzeros uper conducting gaps, even at sm allvalue s of $ \ lambda _{\

rightarrow \Delta_0_$. As_$ \tilde{ \mu }_$ increases,_$_\tilde{ \Delta_}_0_$ reaches a_maximum at an_optimum value of the_chemical potential and_$_\tilde{ \Delta }_0 $ decreases as $ \tilde{ \mu } $ increases even further. For_the_case $_\tilde{_\lambda_}_{ \mbox{\scriptsize{sc}} } < 1_$, our results are shown_in Figure \[FigD0vsmu\_LscLT1\_2\]._Starting at $ \tilde{ \mu } = 0_$,_the system is_in the normal state. However, as $ \tilde{ \mu_} $ increases, the system asymptotically_becomes superconducting and_the_order_parameter also increases up_to a maximum value at an_optimal chemical potential. As $ \tilde{_\mu } $ increases even further, $_\tilde{ \Delta }_0 $ decreases and_the energy gap displays a_dome-shaped plot. Our_results are consistent with [@Fukushima2007],_where chiral and_diquark condensates_are calculated for_two-color and two-flavor QCD. For a_certain choice of_parameters, the authors obtain the same_dispersion_relation given by_Eq. (\[Ej\])_and_their numerical_results for $_\Delta_$ also_display_a dome-shaped plot, as can be_seen_in Fig. 1 of [@Fukushima2007], for the_choice of parameters I,_referred_as the [*weak-coupling case*]{}. Since_the energy gap and the_superconducting critical temperature $ T_{ c_}$ are_proportional, our_results qualitatively reproduce the superconducting phase diagram of several compounds, like_122 pnictides and cuprate superconductors, where_the critical temperature displays_a characteristic_dome_as charge carriers_are_doped into_the system. A dome-like structure of the superconducting_phase for_strongly interacting two-dimensional Dirac fermions has_been previously obtained in_[@Smith2009],_where the authors observed a dome-shaped_plot of the superconducting phase at_intermediate filling fractions, surrounded by_the_normal_phase for fillings close to_unity or zero, which is consistent_to our results. We_conclude this section with a final remark:_contrary_to what happens at $ \mu_=_0 $, where superconductivity appears in_the_system_only when $ \lambda_{_\mbox{\scriptsize{sc}} } > \lambda_{c } $,_at finite chemical potential, we could always find a_finite $ \tilde{_\mu } $ that provided_nonzero_superconducting_gaps, even at small values of $ \lambda_{ \

]. In contrast, the Magnus term aligns the particle velocity perpendicular to the direction of the external forces, causing the skyrmions to move at an angle called the intrinsic skyrmion Hall angle $\theta^{\rm int}_{Sk}$ with respect to the external forces [@2; @10; @12; @13; @14]. As recently shown, the Magnus term strongly affects the overall skyrmion dynamics in the presence of disorder, with the measured skyrmion Hall angle starting at zero or a small value for drives just above depinning and gradually increasing to the intrinsic or pin-free $\theta^{\rm int}_{Sk}$ value as the drive increases and the skyrmions move faster [@12; @13; @14; @15; @16; @20; @21; @22; @23]. In many slowly driven systems with quenched disorder, the motion near depinning takes the form of bursts or avalanches of the type observed in driven magnetic domain walls [@24; @24a; @25], vortices in type-II superconductors [@19; @26], earthquake models [@28], and near yielding transitions in sheared materials [@29; @30]. Avalanches or so-called crackling noise arise in a wide range of collectively interacting driven systems, and scaling properties of the avalanche size distributions as well as the average avalanche shape can be used to determine whether the system is at a nonequilibrium critical point and to identify its universality class [@31; @32; @33]. In many avalanche systems, the dynamics is overdamped, but when non-dissipative effects become important, the statistics of the avalanches can change. In particular, the average avalanche shape becomes asymmetric in the presence of an effective mass or stress overshoots [@25; @32; @33]. An open question is whether skyrmions can exhibit avalanche dynamics and, if so, what impact the Magnus term would have on such dynamics. It is important to understand intermittent skyrmion dynamics near the depinning threshold in order to fully realize applications which require skyrmions to be moved and stopped in a controlled fashion, such as in skyrmion race track memories [@18]. In this work we numerically examine avalanches of slowly driven skyrmions moving over quenched disorder for varied ratios $\alpha_m/\alpha_d$ of the Magnus term to the damping term. When $\alpha_m/\alpha_d

]. In contrast, the Magnus term aligns the particle speed vertical to the focus of the external forces, causing the skyrmions to go at an angle called the intrinsic skyrmion Hall slant $ \theta^{\rm int}_{Sk}$ with respect to the external forces [ @2; @10; @12; @13; @14 ]. As recently shown, the Magnus condition strongly affects the overall skyrmion moral force in the presence of disorder, with the measured skyrmion Hall angle startle at zero or a small value for drives just above depinning and gradually increasing to the intrinsic or pin - free $ \theta^{\rm int}_{Sk}$ value as the campaign increases and the skyrmions move faster [ @12; @13; @14; @15; @16; @20; @21; @22; @23 ]. In many lento driven systems with quelled disorder, the movement near depinning takes the shape of bursts or avalanches of the type note in driven magnetic domain walls [ @24; @24a; @25 ], vortices in character - II superconductors [ @19; @26 ], earthquake model [ @28 ], and near yielding transitions in sheared materials [ @29; @30 ]. Avalanches or so - called crackling noise arise in a across-the-board range of collectively interact drive system, and scaling properties of the avalanche size distribution as well as the average avalanche shape can be used to settle whether the system is at a nonequilibrium critical point and to identify its universality class [ @31; @32; @33 ]. In many avalanche systems, the dynamics is overdamped, but when non - dissipative effect become important, the statistics of the avalanches can change. In especial, the average avalanche form becomes asymmetric in the presence of an effective mass or stress overshoot [ @25; @32; @33 ]. An open question is whether skyrmions can exhibit avalanche dynamics and, if so, what impact the Magnus term would have on such dynamics. It is crucial to understand intermittent skyrmion dynamics near the depinning doorway in order to amply realize lotion which require skyrmions to be act and stopped in a controlled fashion, such as in skyrmion race track memory [ @18 ]. In this work we numerically probe avalanches of slowly driven skyrmions moving over quenched disorderliness for varied ratios $ \alpha_m/\alpha_d$ of the Magnus term to the muffle term. When $ \alpha_m/\alpha_d

]. In contrast, the Magnus tero aligns the pattucle vxlocity perpendkcular to the direction of tie ezternql forces, causing the rkyrmions to move at en angle called vge intrlusic anyrmnoi Hall angle $\thgta^{\rm int}_{Sk}$ fith respect tm ghz external forces [@2; @10; @12; @13; @14]. As recently srown, thr Lagnus term sttonglj wffedns the overall skyrmion dynamica in tht presence of disotder, with the measured skygmioj Hall angle startlng at zero or w small value for drives just above depinning and gradually increaring co the intrunwic mr pin-free $\vheta^{\ri int}_{Sk}$ valuc as tha drive increases and thx sktrmions move faster [@12; @13; @14; @15; @16; @20; @21; @22; @23]. In many slorly drivet aystems with quenxhwd divordar, tfw mutikn nsar deoiniing takes fhe form of bursts or avalanchts jd the type obaerved ig driven magnetic domain walls [@24; @24a; @25], vortpces in type-II superconductirs [@19; @26], earthquake modeps [@28], and nqar yielding transitions in sheared materials [@29; @30]. Aealanrhds ir ri-cwlled crackling noise arise in a wide range os cplkectively intevacting driven sysyels, wnd scaling ptopertiza kf the avalanche slze disjributuons as wtll ax the average avalanche shape can be usvd ti determine whethex the system is st a monequilibrium critical poiht and to ifentify ifr universality cuasx [@31; @32; @33]. In many avalanche systeis, the dyiamicx is ovdrdakped, btt when noj-dissl[ative effects beclme ilpmrtant, the statistics of the avalanches cei change. In psrdicllar, the cveragc avalanche sha[e becomes asykmetric in tfe presencv of an ehfective masf or stress oegrshoots [@25; @32; @33]. Ai open quqstiin iw whethdf skyrmions cam exhibit avalanche dynamics and, if sp, wfzt impact the Mcynow term would hsve on slch drtamics. It is impurtxmt to undersucnd kntetmittent skyrmion dytamids near the depinnonn threshood in orqer to fully tealize applicatiojs whmch reduire skirmions to be moved and stopped in a conhrojled fashion, sucm as in skyrmipn race track memories [@18]. In this work we iumerically examine avaoanches of slowly dtivtn skyrmions kovind over quanched disorder for caried ratios $\alpma_m/\alpha_d$ of the Magnua term to tje damping term. When $\alpha_m/\alpha_d

]. In contrast, the Magnus term aligns velocity to the of the external move an angle called intrinsic skyrmion Hall $\theta^{\rm int}_{Sk}$ with respect to the forces [@2; @10; @12; @13; @14]. As recently shown, the Magnus term strongly the overall skyrmion dynamics in the presence of disorder, with the measured skyrmion angle at or small value for drives just above depinning and gradually increasing to the intrinsic or pin-free $\theta^{\rm value as the drive increases and the skyrmions faster [@12; @13; @14; @16; @20; @21; @22; @23]. many driven systems quenched the near depinning takes form of bursts or avalanches of the type observed in driven magnetic domain walls [@24; @24a; @25], in type-II @26], earthquake [@28], near transitions in sheared @30]. Avalanches or so-called crackling noise wide range of collectively interacting driven systems, and properties of avalanche size distributions as well as average avalanche shape can be used to determine the system is at a nonequilibrium critical point and to identify its universality class [@31; In many avalanche systems, dynamics is overdamped, when effects important, statistics of avalanches can change. In particular, the average avalanche shape becomes asymmetric the presence of an effective mass or stress overshoots [@25; An question is whether can exhibit avalanche dynamics if what impact the Magnus have such important understand skyrmion dynamics near the threshold in order to fully applications which require skyrmions in a controlled fashion, such as in skyrmion track memories [@18]. In this work we examine avalanches of slowly driven skyrmions moving over quenched disorder for varied $\alpha_m/\alpha_d$ of term to the damping term. When $\alpha_m/\alpha_d

]. In contrast, the Magnus term alIgns the parTicle VelOciTy PerpEndiCular to the direCTion Of the external forces, cauSing tHe SKyrmIOnS to moVe at an aNGlE CAllEd ThE inTrINsIc skyRmiOn Hall aNgle $\theta^{\rM inT}_{SK}$ with respect TO tHe external ForCes [@2; @10; @12; @13; @14]. As recentlY shOwn, the maGnuS Term sTroNgly aFfects THe overAll skyrmiOn DYnamicS In the prESEnCe of Disorder, with the meASuREd skyrmion Hall Angle sTaRTiNG At zEro Or a small vaLuE for dRIves jusT AbOVE DepINning and graduAlly increasINg tO the inTrInsIC or pin-Free $\tHeTA^{\rm Int}_{Sk}$ value aS the Drive incrEases aND the skyRMions moVe fastEr [@12; @13; @14; @15; @16; @20; @21; @22; @23]. IN maNy slOWlY dRivEn SYstEMs WitH QueNched disOrDeR, the mOtioN NEAR depInnIng tAkes tHe form of burstS or AvalANchEs of tHe typE obsErVed in Driven MagneTiC domain walls [@24; @24a; @25], voRticEs in type-Ii suPeRcoNdUctorS [@19; @26], EarthqUakE moDels [@28], and Near yieLDinG tRANSiTions in sheared mateRiALS [@29; @30]. AValancheS or so-cALlEd CRackling NoIse ArisE IN a widE ranGE oF collectIvely iNTeRaCting drIvEn systEmS, anD scAling PRopeRties oF the avalAnche SIze distributioNS as well as the aVErAGE aVAlanChe Shape can be uSed tO DeteRminE WhEthER the sYstem Is AT a NOnequilibrium criticAl Point aNd to iDentify its uniVersality cLASS [@31; @32; @33]. In many aValaNChE Systems, the dynaMics iS overdampeD, But when nOn-disSipative Effects beCOMe importAnt, The StaTisTICs Of the avalanchES Can cHaNge. In paRtiCular, thE avEraGe aValAnChe shape bEcomes asYmMeTrIc In tHe preSEnce of an EfFecTiVe mAss or STress oVershOots [@25; @32; @33]. an OpEN quEstion iS WhETHer sKyRmIons Can ExHibit AvalANchE dynamiCs and, if so, WhaT ImpaCt ThE Magnus Term would have On Such dynamiCs. it iS imporTANt to undeRstand intermittent skyrmIOn dynamIcs Near tHe dePinning thResHold in OrdER to fulLy realIze apPlIcaTIOns whICH rEquIrE skyrmions TO Be mOved aNd StopPed in a cOntrolled fashion, suCH as In skyrmion racE trAck mEMOrIes [@18]. iN tHIs wOrK We nUMErically examine Avalanches Of SLoWly driven sKYrmIoNs movinG over quEncheD DisordeR for varieD ratios $\alPhA_m/\alPHA_d$ oF the Magnus Term to thE damping tERm. WheN $\AlPha_m/\aLphA_d

]. In contrast, the Magnus term alig ns th e p art ic le v eloc ity perpendicu l ar t o the direction of the exte rn a l fo r ce s, ca using t h es k yrm io ns to m o ve at a n a ngle ca lled the i ntr in sic skyrmion Ha ll angle $ \th eta^{\rm int }_{ Sk}$ w it h r e spect to theextern a l forc es [@2; @ 10 ; @12;@ 13; @14 ] . A s re cently shown, the Ma g nus term stron gly af fe c ts t heove rall skyrm io n dyn a mics in th e p res e nce of disord er, with th e me asured s kyr m ion Ha ll an gl e st arting at z eroor a smal l valu e for dr i ves jus t abov e d epi nnin g a nd gr ad u all y i ncr e asi ng to th ein trins ic o r p i n-fr ee$\th eta^{ \rm int}_{Sk} $ v alue asthe d riveincr ea ses a nd the skyr mi ons move faster [@1 2; @13; @ 14; @ 15; @ 16; @ 2 0; @21 ; @ 22; @23]. In man y sl ow l y dr iven systems withqu e n ch ed disor der, t h emo t ion near d epi nnin g takes the fo rm of bu rsts o r a va lanches o f thety peobs erved in d rivenmagnetic doma i n walls [@24;@ 24a; @25], vo r ti c e si n ty pe- II supercon duct o rs [ @19; @2 6], earth quake m o de l s [@28], and near y ie ldingtrans itions in she ared mater i a l s [@29;@30] . A v alanches or so -call ed crackli n g noisearise in a wi de rangeo f collect ive lyint era c t in g driven syst e m s, a nd scalin g p roperti esofthe av al anche siz e distri bu ti on saswella s the av er age a val anche shapecan b e us ed t o de termine wh e t herth esyst emis at a non e qui librium critical po i nt a nd t o ident ify its unive rs ality clas s[@3 1; @32 ; @33]. In many avalanche systems , the dy nam ics i s ov erdamped, bu t when no n -dissi pative effe ct s b e c ome i m p or tan t, the stati s t ics of t he ava lanches can change. In pa r tic ular, the ave rag e av a l an che sh a pebe c ome s asymmetric in t he presenc eo fan effecti v e m as s or st ress ov ersho o ts [@25 ; @32; @3 3]. An op en que s t ion is whethe r skyrmi ons can e x hibit av alanc hedynami cs an d, if so, w h atimpac t theMa gnus t erm w ou ld haveon such dynamics. It is impor tanttounderstan d i n ter mittent s kyrm ion dynami csnea r the de p innin g th r es hol d in o rder to fullyr ea liz e ap plicationsw h i chrequi res kyrmio ns t o be moved and st o pped in a cont roll e d fa shi o n, s uc h as in skyrmi onra c e track m em ories [@18] . In th is workwe num erical ly exam i n ea valanc hesofslowly dr ive ns kyrmion smo v ing ov er q ue ncheddisord e r fo r varied ratios $\ alpha _ m /\alp h a_d $ ofth e Magnu s ter m to the d amping term . When $\a lpha_ m/\alph a_ d

]. In_contrast, the_Magnus term aligns the_particle velocity_perpendicular_to the_direction_of the external_forces, causing the_skyrmions to move at_an angle called_the_intrinsic skyrmion Hall angle $\theta^{\rm int}_{Sk}$ with respect to the external forces [@2; @10;_@12;_@13; @14]._As_recently_shown, the Magnus term strongly_affects the overall skyrmion dynamics_in the_presence of disorder, with the measured skyrmion Hall_angle_starting at zero_or a small value for drives just above depinning_and gradually increasing to the intrinsic_or pin-free $\theta^{\rm_int}_{Sk}$_value_as the drive increases_and the skyrmions move faster [@12;_@13; @14; @15; @16; @20; @21;_@22; @23]. In many slowly driven systems with_quenched disorder, the motion near depinning_takes the form of bursts_or avalanches_of the type observed in_driven magnetic domain_walls [@24;_@24a; @25], vortices_in type-II superconductors [@19; @26], earthquake_models [@28], and_near yielding transitions in sheared materials_[@29;_@30]. Avalanches or_so-called_crackling_noise arise_in a wide_range_of collectively_interacting_driven systems, and scaling properties of_the_avalanche size distributions as well as the_average avalanche shape can_be_used to determine whether_the system is at a_nonequilibrium critical point and to identify_its universality_class [@31;_@32; @33]. In many avalanche systems, the dynamics is overdamped, but_when non-dissipative effects become important, the_statistics of the avalanches_can change._In_particular, the average_avalanche_shape becomes_asymmetric in the presence of an effective_mass or_stress overshoots [@25; @32; @33]. An_open question is whether_skyrmions_can exhibit avalanche dynamics and, if_so, what impact the Magnus term_would have on such dynamics._It_is_important to understand intermittent skyrmion_dynamics near the depinning threshold in_order to fully_realize applications which require skyrmions to be_moved_and stopped in a controlled fashion,_such_as in skyrmion race track memories_[@18]. In_this_work we numerically examine avalanches_of slowly driven skyrmions moving over_quenched disorder for varied ratios $\alpha_m/\alpha_d$ of the Magnus_term to the_damping term. When $\alpha_m/\alpha_d

g^{kl}\big[ {\overline{R}}( x_k, x_i, x_l, x_m) h^m_j +{\overline{R}}( x_k, x_i, x_j, x_m) h^m_l + {\overline{R}}(x_m, x_i, x_j, x_l ) h^m_k \\ & + {\overline{R}}(x_m, x_k, x_j, x_l) h^m_i + ({\overline{\nabla}}_{x_l} {\overline{R}})(\nu, x_i, x_j, x_k) + ({\overline{\nabla}}_{ x_j } {\overline{R}})(\nu, x_k, x_i, x_l)\big]. \end{split}$$ Contract with $g^{kl}$ and use the (anti)symmetric properties of the full tensor ${\overline{R}}$. geometry of the warped product {#wp geometry} ------------------------------ The [*warped product*]{}, $P\times_{h} S$, of two Riemannian manifolds, $(P^p,\varpi)$ and $(S^s,\varsigma)$, with [*warping factor*]{} $h$ is the product manifold, $P\times S$, equipped with the metric $$\gamma^{P\times_{h} S} = \varpi \oplus h^2 \varsigma,$$ In this manner, providing $h$ is not degenerate, $(P\times_{h} S, \gamma^{P\times_{h} S})$ also becomes a Riemannian manifold.\ \ In this section we look at some properties of warped product spaces in a more specialised setting. We consider the warped product $I\times_f M^n$, where $I$ is an arbitrary one dimensional manifold parameterised with $r$, $(M^n,\sigma)$ is a Riemannian $n$-manifold with metric $\sigma$, and $f$ is a positive monotonically increasing (preferably differentiable) function of $r$. Thus we have the metric $$\label{wp metric} \gamma =\operatorname{d}\!r^2 \oplus f^2 \sigma.$$ Indeed, this metric is conformal to a metric of the form $$\tag{*}\label{metric} \operatorname{d}\!\rho^2 \oplus \

g^{kl}\big [ { \overline{R } } (x_k, x_i, x_l, x_m) h^m_j + { \overline{R } } (x_k, x_i, x_j, x_m) h^m_l + { \overline{R}}(x_m, x_i, x_j, x_l) h^m_k \\ & + { \overline{R}}(x_m, x_k, x_j, x_l) h^m_i + ({ \overline{\nabla}}_{x_l } { \overline{R}})(\nu, x_i, x_j, x_k) + ({ \overline{\nabla } } _ { x_j } { \overline{R}})(\nu, x_k, x_i, x_l)\big ]. \end{split}$$ Contract with $ g^{kl}$ and use the (anti)symmetric properties of the full tensor $ { \overline{R}}$. geometry of the warped product { # wp geometry } ------------------------------ The [ * warped merchandise * ] { }, $ P\times_{h } S$, of two Riemannian manifold, $ (P^p,\varpi)$ and $ (S^s,\varsigma)$, with [ * warping factor * ] { } $ h$ is the product manifold, $ P\times S$, equipped with the measured $ $ \gamma^{P\times_{h } S } = \varpi \oplus h^2 \varsigma,$$ In this manner, providing $ h$ is not degenerate, $ (P\times_{h } S, \gamma^{P\times_{h } S})$ also become a Riemannian manifold.\ \ In this section we look at some property of warped product spaces in a more specialized setting. We consider the warped product $ I\times_f M^n$, where $ I$ is an arbitrary one dimensional manifold parameterised with $ r$, $ (M^n,\sigma)$ is a Riemannian $ n$-manifold with measured $ \sigma$, and $ f$ is a positive monotonically increasing (preferably differentiable) affair of $ r$. Thus we have the measured $ $ \label{wp metric } \gamma = \operatorname{d}\!r^2 \oplus f^2 \sigma.$$ Indeed, this metric is conformal to a metric of the shape $ $ \tag{*}\label{metric } \operatorname{d}\!\rho^2 \oplus \

g^{kp}\big[ {\overline{R}}( x_k, x_i, x_l, w_m) h^m_j +{\overline{T}}( z_k, x_i, e_j, x_m) h^j_l + {\overuine{R}}(x_m, x_i, x_j, x_l ) h^m_k \\ & + {\overline{R}}(x_m, x_y, x_j, x_l) h^l_i + ({\overoine{\iabla}}_{x_l} {\overline{C}})(\hu, x_i, x_m, x_k) + ({\lvernmne{\nabla}}_{ x_j } {\ovgrline{R}})(\nu, x_k, x_i, x_l)\big]. \eud{split}$$ Contract with $g^{kl}$ and use the (anti)sykmftric propertigs of ehe rllo tensor ${\overline{R}}$. geometry of the wagped product {#wp grometry} ------------------------------ The [*warped product*]{}, $O\timfs_{h} S$, of two Riemajnian manifilds, $(P^p,\varpi)$ and $(R^s,\varsigma)$, with [*warpinf factor*]{} $h$ is the product manifuld, $P\cimes S$, equupped fith the mevric $$\gwmma^{P\times_{h} S} = \varpi \oplus h^2 \varsigma,$$ In thms mqnner, providing $h$ is iot degenerate, $(P\timef_{h} S, \gammd^{P\cimes_{h} S})$ also becomes a Riemdnnidn mxbifuld.\ \ Jn tgis seftikn we look at some priperties of warped kroqlvt spaces in a morq fpecialised setting. We consider the warked pdoduct $I\times_f M^n$, where $I$ is an arbitrary ong dimensiogal manifold parameterised with $r$, $(M^n,\sigma)$ is a Riamannmav $n$-nakifoue aith metric $\sigma$, and $f$ is a positive monotonydaklj increasing (prefcrably differentianlf) gonction of $r$. Tfus we habe the metric $$\labep{wp mettic} \gamma =\optratotname{d}\!r^2 \oplus f^2 \sigma.$$ Indeee, this metrib is conformal to a mecric of the yorm $$\tsg{*}\labrl{metric} \operatornaoe{d}\!\rgo^2 \oplus \

g^{kl}\big[ {\overline{R}}( x_k, x_i, x_l, x_m) h^m_j x_i, x_m) h^m_l {\overline{R}}(x_m, x_i, x_j, + x_k, x_j, x_l) + ({\overline{\nabla}}_{x_l} {\overline{R}})(\nu, x_j, x_k) + ({\overline{\nabla}}_{ x_j } x_k, x_i, x_l)\big]. \end{split}$$ Contract with $g^{kl}$ and use the (anti)symmetric properties of full tensor ${\overline{R}}$. geometry of the warped product {#wp geometry} ------------------------------ The [*warped $P\times_{h} of Riemannian $(P^p,\varpi)$ and $(S^s,\varsigma)$, with [*warping factor*]{} $h$ is the product manifold, $P\times S$, equipped with the $$\gamma^{P\times_{h} S} = \varpi \oplus h^2 \varsigma,$$ In manner, providing $h$ is degenerate, $(P\times_{h} S, \gamma^{P\times_{h} S})$ becomes Riemannian manifold.\ In section look at some of warped product spaces in a more specialised setting. We consider the warped product $I\times_f M^n$, where is an dimensional manifold with $(M^n,\sigma)$ a Riemannian $n$-manifold $\sigma$, and $f$ is a positive differentiable) function of $r$. Thus we have the $$\label{wp metric} =\operatorname{d}\!r^2 \oplus f^2 \sigma.$$ Indeed, this is conformal to a metric of the form \operatorname{d}\!\rho^2 \oplus \

g^{kl}\big[ {\overline{R}}( x_k, x_i, x_l, x_m) h^m_J +{\overline{R}}( X_k, x_i, x_J, x_m) H^m_l + {\OvErliNe{R}}(x_M, x_i, x_j, x_l ) h^m_k \\ & + {\overLIne{R}}(X_m, x_k, x_j, x_l) h^m_i + ({\overline{\nabLa}}_{x_l} {\oVeRLine{r}})(\Nu, X_i, x_j, x_K) + ({\overliNE{\nABLa}}_{ x_J } {\oVeRliNe{r}})(\Nu, X_k, x_i, x_L)\biG]. \end{splIt}$$ Contract WitH $g^{Kl}$ and use the (aNTi)Symmetric pRopErties of the fUll Tensor ${\OvErlINe{R}}$. geOmeTry of The warPEd prodUct {#wp geomEtRY} ------------------------------ The [*waRPed prodUCT*]{}, $P\TimeS_{h} S$, of two RiemanniaN MaNIfolds, $(P^p,\varpi)$ aNd $(S^s,\vaRsIGmA)$, WIth [*WarPing factor*]{} $H$ iS the pROduct maNIfOLD, $p\tiMEs S$, equipped wiTh the metric $$\GAmmA^{P\timeS_{h} s} = \vaRPi \opluS h^2 \varSiGMa,$$ IN this manner, ProvIding $h$ is nOt degeNErate, $(P\tIMes_{h} S, \gaMma^{P\tiMes_{H} S})$ aLso bECoMeS a RIeMAnnIAn ManIFolD.\ \ In this sEcTiOn we lOok aT SOME proPerTies Of warPed product spaCes In a mORe sPeciaLised SettInG. We coNsider The waRpEd product $I\times_F M^n$, wHere $I$ is an ArbItRarY oNe dimENsionaL maNifOld paraMeterisED wiTh $R$, $(m^N,\SiGma)$ is a Riemannian $n$-mAnIFOlD with metRic $\sigMA$, aNd $F$ Is a positIvE moNotoNICally IncrEAsIng (prefeRably dIFfErEntiablE) fUnctioN oF $r$. THus We havE The mEtric $$\lAbel{wp meTric} \gAMma =\operatornamE{D}\!r^2 \oplus f^2 \sigma.$$ iNdEED, tHIs meTriC is conformaL to a MEtriC of tHE fOrm $$\TAg{*}\labEl{metRiC} \OpERatorname{d}\!\rho^2 \oplus \

g^{kl}\big[ {\overline{R} }( x_k, x_ i, x_ l,x_m )h^m_ j + {\overline{R}} ( x_k , x_i, x_j, x_m) h^m_l + {\ ov e rlin e {R }}(x_ m, x_i, x_ j , x_ l)h^m _k \\ & +{\o ve rline{R}}(x_ m ,x_k, x_j,x_l ) h^m_i + ({ \ov erline {\ nab l a}}_{ x_l } {\o verlin e {R}})( \nu, x_i, x _ j, x_k ) + ({\o v e rl ine{ \nabla}}_{ x_j }{ \o v erline{R}})(\n u, x_k ,x _i , x_l )\b ig]. \end { split}$ $ C o n tra c t with $g^{kl }$ and uset he(anti) sy mme t ric pr opert ie s of the full t enso r ${\over line{R } }$. ge o metry o f thewar ped pro d uc t{#w pg eom e tr y}- --- -------- -- -- ----- ---- - - - - - T he[*wa rpedproduct*]{},$P\ time s _{h } S$, of t wo R ie manni an man ifold s, $(P^p,\varpi)$ and $(S^s,\v ars ig ma) $, with [*warp ing fa ctor*]{ } $h$ i s th ep r o du ct manifold, $P\ti me s S$ , equipp ed wit h t he metric $ $\ gam ma^{ P \ times _{h} S} = \varp i \opl u s h ^2 \var si gma,$$ I n t his mann e r, p rovidi ng $h$ i s not degenerate, $( P \times_{h} S, \g a m ma ^ {P\t ime s_{h} S})$also beco mesa R iem a nnian mani fo l d. \ \ In this sectionwe lookat so me properties of warped p r oduct sp aces in a more special isedsetting. W e conside r the warpedproduct $ I \ times_fM^n $,whe re$ I $is an arbitra r y one d imensio nal manifo ldpar ame ter is ed with $ r$, $(M^ n, \s ig ma )$is aR iemannia n$n$ -m ani foldw ith me tric$\si gm a$ , an d $f$ i s a p osit iv emono ton ic allyincr e asi ng (pre ferably d iff e rent ia bl e) func tion of $r$.Th us we have t hemetric $ $\label{ wp metric} \gamma = \ operato rna me{d} \!r^ 2 \oplusf^2 \sigm a.$ $ Indee d, thi s met ri c i s confo r m al to a metric of t heform$$ \tag {*}\lab el{metric} \o p era torname{d}\!\ rho ^2 \ o p lu s \

g^{kl}\big[_{\overline{R}}( x_k,_x_i, x_l, x_m) h^m_j_ +{\overline{R}}(_x_k,_x_i, x_j,_x_m)_h^m_l + {\overline{R}}(x_m,_x_i, x_j, x_l_) h^m_k \\ _ __ _&_+ {\overline{R}}(x_m,_x_k,_x_j,_x_l) h^m_i + ({\overline{\nabla}}_{x_l} {\overline{R}})(\nu,_x_i, x_j, x_k) + ({\overline{\nabla}}_{_x_j }_{\overline{R}})(\nu, x_k, x_i, x_l)\big]. __ \end{split}$$ Contract_with $g^{kl}$ and use the (anti)symmetric properties of the_full tensor ${\overline{R}}$. geometry of the warped_product {#wp geometry} ------------------------------ The_[*warped_product*]{},_$P\times_{h} S$, of two_Riemannian manifolds, $(P^p,\varpi)$ and $(S^s,\varsigma)$, with_[*warping factor*]{} $h$ is the product_manifold, $P\times S$, equipped with the metric_$$\gamma^{P\times_{h} S} = \varpi \oplus _h^2 \varsigma,$$ In this manner,_providing $h$_is not degenerate, $(P\times_{h} S,_\gamma^{P\times_{h} S})$ also_becomes a_Riemannian manifold.\ \ In this_section we look at some properties_of warped product_spaces in a more specialised setting._We_consider the warped_product_$I\times_f_M^n$, where_$I$ is an_arbitrary_one dimensional_manifold_parameterised with $r$, $(M^n,\sigma)$ is a_Riemannian_$n$-manifold with metric $\sigma$, and $f$ is_a positive monotonically increasing_(preferably_differentiable) function of $r$._Thus we have the metric_$$\label{wp metric} \gamma_=\operatorname{d}\!r^2 \oplus_f^2 \sigma.$$_Indeed, this metric is conformal to a metric of the form_$$\tag{*}\label{metric} \operatorname{d}\!\rho^2_\oplus \

& 1.18 & 0.79, 2.16 & 5.18\ [\[[Ne]{}\]]{}15.55/[\[[Ne]{}\]]{}24.32& 4.53 & 0.79, 0.75 & 5.15\ [\[[Ne]{}\]]{}7.65/[\[[Ne]{}\]]{}15.55& 7.29 & 1.70, 0.79 & 5.34\ [\[[Ne]{}\]]{}12.81/[\[[Ne]{}\]]{}14.32& 0.99 & 0.40, 2.16 & 5.08\ [lccr]{} & Term & $\lambda (\!{\mbox{$\,\mu$m}})$ & $n_{\rm c}$ (cm$^{-3}$)\ [\[[Si]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 1.4290 & $1.0\times10^{10}$\ [\[[Si]{}\]]{} & $^2$P$^o_{1/2}-^2$P$^o_{3/2}$ & 1.9625 & $1.7\times10^9$\ [\[[Si]{}\]]{} & $^3$P$_0-^3$P$_2$ & 2.4822 & $7.0\times10^8$\ [\[[Mg]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 3.0269 & $2.6\times10^8$\ [\[[S]{}\]]{} & $^3$P$_0-^3$P$_1$ & 3.7495 & $4.5\times10^8$\ [\[[Ne]{}\]]{} & $^2$P$_{1/2}-^2$P$_{1/3}$ & 12.808 & $2.8\times10^6$\ [\[[Ne]{}\]]{} & $^3$P$_2-^3$P$_1$ & 14.311 & $1.4\times10^5$\ [\[[Ne]{}\]]{} & $^3$P$_1-^3$P$_2$ & 15.550 & $8.1\times10^5$\ [\[[Ne]{}\]]{} & $^3$P$_1-^3$P$_0$ &

& 1.18 & 0.79, 2.16 & 5.18\ [ \[[Ne]{}\]]{}15.55/[\[[Ne]{}\]]{}24.32 & 4.53 & 0.79, 0.75 & 5.15\ [ \[[Ne]{}\]]{}7.65/[\[[Ne]{}\]]{}15.55 & 7.29 & 1.70, 0.79 & 5.34\ [ \[[Ne]{}\]]{}12.81/[\[[Ne]{}\]]{}14.32 & 0.99 & 0.40, 2.16 & 5.08\ [ lccr ] { } & Term & $ \lambda (\!{\mbox{$\,\mu$m}})$ & $ n_{\rm c}$ (cm$^{-3}$)\ [ \[[Si]{}\ ] ] { } & $ ^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 1.4290 & $ 1.0\times10^{10}$\ [ \[[Si]{}\ ] ] { } & $ ^2$P$^o_{1/2}-^2$P$^o_{3/2}$ & 1.9625 & $ 1.7\times10 ^ 9$\ [ \[[Si]{}\ ] ] { } & $ ^3$P$_0-^3$P$_2 $ & 2.4822 & $ 7.0\times10 ^ 8$\ [ \[[Mg]{}\ ] ] { } & $ ^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 3.0269 & $ 2.6\times10 ^ 8$\ [ \[[S]{}\ ] ] { } & $ ^3$P$_0-^3$P$_1 $ & 3.7495 & $ 4.5\times10 ^ 8$\ [ \[[Ne]{}\ ] ] { } & $ ^2$P$_{1/2}-^2$P$_{1/3}$ & 12.808 & $ 2.8\times10 ^ 6$\ [ \[[Ne]{}\ ] ] { } & $ ^3$P$_2-^3$P$_1 $ & 14.311 & $ 1.4\times10 ^ 5$\ [ \[[Ne]{}\ ] ] { } & $ ^3$P$_1-^3$P$_2 $ & 15.550 & $ 8.1\times10 ^ 5$\ [ \[[Ne]{}\ ] ] { } & $ ^3$P$_1-^3$P$_0 $ &

& 1.18 & 0.79, 2.16 & 5.18\ [\[[Ne]{}\]]{}15.55/[\[[Ne]{}\]]{}24.32& 4.53 & 0.79, 0.75 & 5.15\ [\[[Ne]{}\]]{}7.65/[\[[Ne]{}\]]{}15.55& 7.29 & 1.70, 0.79 & 5.34\ [\[[Nt]{}\]]{}12.81/[\[[Ne]{}\]]{}14.32& 0.99 & 0.40, 2.16 & 5.08\ [lccr]{} & Tetm & $\lambva (\!{\mbox{$\,\ju$m}})$ & $n_{\rm c}$ (cm$^{-3}$)\ [\[[Si]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 1.4290 & $1.0\times10^{10}$\ [\[[Si]{}\]]{} & $^2$P$^o_{1/2}-^2$'$^o_{3/2}$ & 1.9625 & $1.7\timtf10^9$\ [\[[Si]{}\]]{} & $^3$P$_0-^3$P$_2$ & 2.4822 & $7.0\times10^8$\ [\[[Mg]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 3.0269 & $2.6\times10^8$\ [\[[S]{}\]]{} & $^3$P$_0-^3$P$_1$ & 3.7495 & $4.5\times10^8$\ [\[[Ne]{}\]]{} & $^2$P$_{1/2}-^2$P$_{1/3}$ & 12.808 & $2.8\tlies10^6$\ [\[[Ns]{}\]]{} & $^3$P$_2-^3$P$_1$ & 14.311 & $1.4\times10^5$\ [\[[Ne]{}\]]{} & $^3$P$_1-^3$P$_2$ & 15.550 & $8.1\times10^5$\ [\[[Na]{}\]]{} & $^3$P$_1-^3$P$_0$ &

& 1.18 & 0.79, 2.16 & 5.18\ & 0.75 & [\[[Ne]{}\]]{}7.65/[\[[Ne]{}\]]{}15.55& 7.29 & 0.99 0.40, 2.16 & [lccr]{} & Term $\lambda (\!{\mbox{$\,\mu$m}})$ & $n_{\rm c}$ (cm$^{-3}$)\ & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 1.4290 & $1.0\times10^{10}$\ [\[[Si]{}\]]{} & $^2$P$^o_{1/2}-^2$P$^o_{3/2}$ & 1.9625 & $1.7\times10^9$\ & $^3$P$_0-^3$P$_2$ & 2.4822 & $7.0\times10^8$\ [\[[Mg]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 3.0269 & $2.6\times10^8$\ & & & [\[[Ne]{}\]]{} & $^2$P$_{1/2}-^2$P$_{1/3}$ & 12.808 & $2.8\times10^6$\ [\[[Ne]{}\]]{} & $^3$P$_2-^3$P$_1$ & 14.311 & $1.4\times10^5$\ [\[[Ne]{}\]]{} & & 15.550 & $8.1\times10^5$\ [\[[Ne]{}\]]{} & $^3$P$_1-^3$P$_0$ &

& 1.18 & 0.79, 2.16 & 5.18\ [\[[Ne]{}\]]{}15.55/[\[[Ne]{}\]]{}24.32& 4.53 & 0.79, 0.75 & 5.15\ [\[[Ne]{}\]]{}7.65/[\[[Ne]{}\]]{}15.55& 7.29 & 1.70, 0.79 & 5.34\ [\[[Ne]{}\]]{}12.81/[\[[Ne]{}\]]{}14.32& 0.99 & 0.40, 2.16 & 5.08\ [lccr]{} & Term & $\lambdA (\!{\mbox{$\,\mu$m}})$ & $n_{\rM c}$ (cm$^{-3}$)\ [\[[SI]{}\]]{} & $^2$P$^o_{3/2}-^2$p$^o_{1/2}$ & 1.4290 & $1.0\tImEs10^{10}$\ [\[[Si]{}\]]{} & $^2$p$^o_{1/2}-^2$P$^o_{3/2}$ & 1.9625 & $1.7\Times10^9$\ [\[[Si]{}\]]{} & $^3$P$_0-^3$P$_2$ & 2.4822 & $7.0\times10^8$\ [\[[mG]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^O_{1/2}$ & 3.0269 & $2.6\times10^8$\ [\[[S]{}\]]{} & $^3$P$_0-^3$P$_1$ & 3.7495 & $4.5\times10^8$\ [\[[Ne]{}\]]{} & $^2$P$_{1/2}-^2$P$_{1/3}$ & 12.808 & $2.8\timeS10^6$\ [\[[Ne]{}\]]{} & $^3$P$_2-^3$P$_1$ & 14.311 & $1.4\TiMEs10^5$\ [\[[Ne]{}\]]{} & $^3$p$_1-^3$p$_2$ & 15.550 & $8.1\tImes10^5$\ [\[[NE]{}\]]{} & $^3$P$_1-^3$P$_0$ &

& 1.18 & 0.79, 2.16 & 5.18 \ [\[[Ne]{ }\]]{ }15 .55 /[ \[[N e]{} \]]{}24.32& 4. 5 3 &0.79, 0.75 & 5.15\ [\[ [Ne]{ }\ ] ]{}7 . 65 /[\[[ Ne]{}\] ] {} 1 5 .55 &7. 29&1 .7 0, 0. 79& 5.34\ [\[[Ne]{} \]] {} 12.81/[\[[Ne ] {} \]]{}14.32 & 0 .99 & 0.40,2.1 6 & 5. 08 \ [ lccr] {}& Ter m & $\ l ambda(\!{\mbox {$ \ ,\mu$m } })$ & $ n _ {\ rm c }$ (cm$^{-3}$)\ [ \ [[ S i]{}\]]{} & $^ 2$P$^o _{ 3 /2 } - ^2$ P$^ o_{1/2}$ & 1 .4290 & $1.0\ t im e s 1 0^{ 1 0}$\ [\[[Si]{ }\]]{} & $^ 2 $P$ ^o_{1/ 2} -^2 $ P$^o_{ 3/2}$ & 1.9 625 & $1.7\ time s10^9$\ [ \[[Si] { }\]]{}& $^3$P$ _0-^3$ P$_ 2$& 2. 4 82 2& $ 7. 0 \ti m es 10^ 8 $\[\[[Mg]{ }\ ]] {} &$^2$ P $ ^ o _{3/ 2}- ^2$P $^o_{ 1/2}$ & 3.026 9 & $2. 6 \ti mes10 ^8$\[\[[ S] {}\]] {} & $ ^3$P$ _0 -^3$P$_1$ & 3.7 495& $4.5\ti mes 10 ^8$ \[\[[N e ]{}\]] {}& $ ^2$P$_{ 1/2}-^2 $ P$_ {1 / 3 } $& 12.808 & $2.8\ti me s 1 0^ 6$\ [\[[ Ne]{}\ ] ]{ }& $^3$P$_ 2- ^3$ P$_1 $ & 14. 311& $ 1.4\time s10^5$ \ [ \[ [Ne]{}\ ]] {} & $ ^3 $P$ _1- ^3$P$ _ 2$ & 15.55 0 & $8.1 \time s 10^5$\ [\[[Ne] { }\]]{} & $^3$ P $_ 1 - ^3 $ P$_0 $ &

& 1.18_& 0.79,_2.16 & 5.18\ [\[[Ne]{}\]]{}15.55/[\[[Ne]{}\]]{}24.32& 4.53_& 0.79,_0.75_& 5.15\ [\[[Ne]{}\]]{}7.65/[\[[Ne]{}\]]{}15.55&_7.29_& 1.70, 0.79_& 5.34\ [\[[Ne]{}\]]{}12.81/[\[[Ne]{}\]]{}14.32& 0.99_& 0.40, 2.16 &_5.08\ [lccr]{} & Term_&_$\lambda (\!{\mbox{$\,\mu$m}})$ & $n_{\rm c}$ (cm$^{-3}$)\ [\[[Si]{}\]]{} & $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 1.4290 & $1.0\times10^{10}$\ [\[[Si]{}\]]{} & $^2$P$^o_{1/2}-^2$P$^o_{3/2}$_&_1.9625 &_$1.7\times10^9$\ [\[[Si]{}\]]{}_&_$^3$P$_0-^3$P$_2$ & 2.4822 & $7.0\times10^8$\ [\[[Mg]{}\]]{}_& $^2$P$^o_{3/2}-^2$P$^o_{1/2}$ & 3.0269 &_$2.6\times10^8$\ [\[[S]{}\]]{} &_$^3$P$_0-^3$P$_1$ & 3.7495 & $4.5\times10^8$\ [\[[Ne]{}\]]{} & $^2$P$_{1/2}-^2$P$_{1/3}$ &_12.808_& $2.8\times10^6$\ [\[[Ne]{}\]]{} &_$^3$P$_2-^3$P$_1$ & 14.311 & $1.4\times10^5$\ [\[[Ne]{}\]]{} & $^3$P$_1-^3$P$_2$ & 15.550_& $8.1\times10^5$\ [\[[Ne]{}\]]{} & $^3$P$_1-^3$P$_0$ &

Tlidi, M. Georgiou, and P. Mandel, Phys. Rev. A **48**, 4605 (1993); M. Tlidi, P. Mandel, and R. Lefever, Phys. Rev. Lett. **73**, 640 (1994). V. J. Sánchez-Morcillo and G. J. de Valcárcel, Quantum Semiclass. Opt. **8**, 919 (1996). A. H. Nayfeh, *Perturbation methods* (John Wiley & Sons, New York, 2000) G. Kozyreff, S. J. Chapman, and M. Tlidi, Phys. Rev. E** 68**, 015201(R) (2003); G. Kozyreff and M. Tlidi, Phys. Rev. E **69**, 066202 (2004). C. Durniak *et al.*, Phys. Rev. E** 72**, 026607 (2005). M. G. Clerc, A. Petrossian, and S. Residori, Phys. Rev. E, **71,** 015205 (R) (2005). --- abstract: 'We investigate the numerical implementation of functionally graded properties in the context of the finite element method. The macroscopic variation of elastic properties inherent to functionally graded materials (FGMs) is introduced at the element level by means of the two most commonly used schemes: (i) nodal based gradation, often via an auxiliary (non-physical) temperature-dependence, and (ii) Gauss integration point based gradation. These formulations are extensively compared by solving a number of paradigmatic boundary value problems for which analytical solutions can be obtained. The nature of the notable differences revealed by the results is investigated in detail. We provide a user subroutine for the finite element package ABAQUS to overcome the limitations of the most popular approach for implementing FGMs in commercial software. The use of reliable, element-based formulations to define the material property variation could be key in fracture assessment of FGMs and other non-homogeneous materials.' address: 'Department of Engineering, University of Cambridge, CB2 1PZ Cambridge, UK' author: - 'Emilio Martínez-Pañeda' bibliography: - 'library.bib' title: On the Finite Element Implementation of

Tlidi, M. Georgiou, and P. Mandel, Phys. Rev. A * * 48 * *, 4605 (1993); M. Tlidi, P. Mandel, and R. Lefever, Phys. Rev. Lett. * * 73 * *, 640 (1994). V. J. Sánchez - Morcillo and G. J. de Valcárcel, Quantum Semiclass. Opt. * * 8 * *, 919 (1996). A. H. Nayfeh, * Perturbation methods * (John Wiley & Sons, New York, 2000) G. Kozyreff, S. J. Chapman, and M. Tlidi, Phys. Rev. E * *   68 * *, 015201(R) (2003); G. Kozyreff and M. Tlidi, Phys. Rev. E * * 69 * *, 066202 (2004). C. Durniak * et al. *, Phys. Rev. E * *   72 * *, 026607 (2005). M.   G. Clerc, A. Petrossian, and S. Residori, Phys. Rev. E, * * 71, * * 015205 (R) (2005). --- abstract:' We investigate the numerical execution of functionally grade properties in the context of the finite element method acting. The macroscopic variation of elastic property inherent to functionally grade material (FGMs) is introduced at the component level by means of the two about commonly used schemes: (i) nodal free-base gradation, often via an auxiliary (non - physical) temperature - dependence, and (ii) Gauss consolidation point based gradation. These conceptualization are extensively compared by solving a act of paradigmatic boundary value problems for which analytic solutions can be obtained. The nature of the notable differences revealed by the result is investigated in detail. We provide a user subroutine for the finite element package ABAQUS to overcome the limitations of the most popular approach for implementing FGMs in commercial software. The use of reliable, element - based formulation to define the material place mutant could be cardinal in fracture assessment of FGMs and other non - homogeneous materials.' address:' Department of Engineering, University of Cambridge, CB2 1PZ Cambridge, UK' author: -' Emilio Martínez - Pañeda' bibliography: -' library.bib' title: On the Finite Element Implementation of

Tlldi, M. Georgiou, and P. Manael, Phys. Rev. A **48**, 4605 (1993); M. Tlmdi, P. Mzndel, ana R. Lefever, Phys. Rev. Lett. **73**, 640 (1994). T. J. Wánchee-Iorcillo and G. J. dd Valcárcep, Quantun Senuclass. Opt. **8**, 919 (1996). A. H. Nayfeh, *Lcrturyavion methods* (Jomn Wiley & Smns, New York, 2000) G. Kuzvreff, S. J. Chapman, and M. Tlidi, Phys. Rqv. E** 68**, 015201(R) (2003); G. Kozyreff and I. Tlpdy, Phgs. Rev. E **69**, 066202 (2004). C. Durniak *et al.*, Phys. Reb. E** 72**, 026607 (2005). M. G. Clerc, A. Petrpssian, and S. Residori, Phys. Rev. E, **71,** 015205 (R) (2005). --- abstract: 'Wf investigajs trw numerical kmplementauimn of funcjionally graded properties in thd concext of the funihg element mevhod. Tre macroscoplb variadion of elastic propevties inyerent to functionallb graded materials (FDMs) is indrkduced at the elenebt leeel ty mdqns of tie fwo modt rommonly ussd schemes: (u) nodal based gradauiog, often via an auxilyawy (non-physical) temperature-dependence, ang (ij) Gauss integration poibt based gradation. Thgse formulwtions are extensively compared by solving a numbar of oarceigmaguc boundary value problems for which analytical aokunions can be obtalned. The nature of tje gotable diffetences xsvsaled by the resulhs is igvestugated in detsil. We provide a user subroytine for thv fibite element packaye ABAQUS to ovetcome yhe limitations of the oost popular aporoach fod implementing FGOs pn cmmmercial software. The use of reliaule, ekement-bxsed formujations to defika the material prooerty vdriation cluld be key in fracture assessmxit of FGMs anc mthvr non-homjgenepus materials.' wddress: 'Departkent of Engiveering, Unpversity mf Cambridgq, CB2 1PZ Cambrhfge, UK' author: - 'Emilio Iartíbez-Pqñeda' bibukography: - 'libraty.bib' titlv: Jn tye Finite Element Lmpleosntation of

Tlidi, M. Georgiou, and P. Mandel, Phys. **48**, (1993); M. P. Mandel, and **73**, (1994). V. J. and G. J. Valcárcel, Quantum Semiclass. Opt. **8**, 919 A. H. Nayfeh, *Perturbation methods* (John Wiley & Sons, New York, 2000) G. S. J. Chapman, and M. Tlidi, Phys. Rev. E** 68**, 015201(R) (2003); G. and Tlidi, Rev. **69**, 066202 (2004). C. Durniak *et al.*, Phys. Rev. E** 72**, 026607 (2005). M. G. Clerc, Petrossian, and S. Residori, Phys. Rev. E, **71,** (R) (2005). --- abstract: investigate the numerical implementation of graded in the of finite method. The macroscopic of elastic properties inherent to functionally graded materials (FGMs) is introduced at the element level by means the two used schemes: nodal gradation, via an auxiliary and (ii) Gauss integration point based are extensively compared by solving a number of boundary value for which analytical solutions can be The nature of the notable differences revealed by results is investigated in detail. We provide a user subroutine for the finite element package overcome the limitations of most popular approach implementing in software. use of element-based formulations to define the material property variation could be key fracture assessment of FGMs and other non-homogeneous materials.' address: 'Department University Cambridge, CB2 1PZ UK' author: - 'Emilio bibliography: 'library.bib' title: On the Implementation

Tlidi, M. Georgiou, and P. Mandel, PHys. Rev. A **48**, 4605 (1993); M. TlIdi, P. MAndEl, aNd r. LefEver, phys. Rev. Lett. **73**, 640 (1994). V. J. SÁNcheZ-Morcillo and G. J. de ValcárCel, QuAnTUm SeMIcLass. OPt. **8**, 919 (1996). A. H. NayFEh, *pERtuRbAtIon MeTHoDs* (JohN WiLey & Sons, new York, 2000) G. KoZyrEfF, S. J. Chapman, anD m. TLidi, Phys. ReV. E** 68**, 015201(R) (2003); g. Kozyreff and m. TlIdi, PhyS. REv. E **69**, 066202 (2004). c. durniAk *eT al.*, PhYs. Rev. E** 72**, 026607 (2005). m. g. Clerc, a. PetrossiAn, ANd S. ResIDori, PhyS. rEv. e, **71,** 015205 (R) (2005). --- abStract: 'We investigaTE tHE numerical implEmentaTiON oF FUncTioNally gradeD pRoperTIes in thE CoNTEXt oF The finite elemEnt method. ThE MacRoscopIc VarIAtion oF elasTiC ProPerties inheRent To functioNally gRAded matERials (FGms) is inTroDucEd at THe ElEmeNt LEveL By MeaNS of The two moSt CoMmonlY useD SCHEmes: (I) noDal bAsed gRadation, often Via An auXIliAry (noN-physIcal) TeMperaTure-dePendeNcE, and (ii) Gauss inteGratIon point bAseD gRadAtIon. ThESe formUlaTioNs are exTensiveLY coMpARED bY solving a number of pArADIgMatic bouNdary vALuE pROblems foR wHicH anaLYTical SoluTIoNs can be oBtaineD. thE nAture of ThE notabLe DifFerEnces REveaLed by tHe resultS is inVEstigated in detAIl. We provide a uSEr SUBrOUtinE foR the finite eLemeNT pacKage abAqUS TO overCome tHe LImITations of the most popUlAr apprOach fOr implementinG FGMs in comMERCial softWare. tHe USe of reliable, elEment-Based formuLAtions to DefinE the mateRial propeRTY variatiOn cOulD be Key IN FrActure assessmENT of FgMS and othEr nOn-homogEneOus MatEriAlS.' address: 'DEpartmenT oF ENgInEerIng, UnIVersity oF CAmbRiDge, cB2 1PZ CAMbridgE, UK' auThor: - 'emIlIO MaRtínez-PAÑeDA' BiblIoGrAphy: - 'LibRaRy.bib' TitlE: on tHe FinitE Element IMplEMentAtIoN of

Tlidi, M. Georgiou, and P . Mandel,Phys. Re v.A**48 **,4605 (1993); M . Tli di, P. Mandel, and R.Lefev er , Phy s .Rev.Lett. * * 73 * * , 6 40 ( 199 4) . V. J. Sá nchez-M orcillo an d G .J. de Valcár c el , QuantumSem iclass. Opt. ** 8**, 9 19 (1 9 96). A. H. N ayfeh, *Pertu rbation m et h ods* ( J ohn Wil e y & Son s, New York, 2000 ) G . Kozyreff, S. J. Ch ap m an , and M. Tlidi, Ph ys . Rev . E** 68 * *, 0 1 520 1 (R) (2003); G . Kozyreffa ndM. Tli di , P h ys. Re v. E** 6 9** , 066202 (2 004) . C. Dur niak * e t al.*, Phys. R ev. E* * 7 2** , 02 6 60 7(20 05 ) . M .G.C ler c, A. Pe tr os sian, and S . Resi dor i, P hys.Rev. E, **71, **0152 0 5 ( R) (2 005). -- -abstr act: ' We in ve stigate the num eric al implem ent at ion o f fun c tional lygra ded pro perties inth e c on text of the finite e l e me nt metho d. The ma cr o scopic v ar iat iono f elas ticp ro pertiesinhere n tto functi on ally g ra ded ma teria l s (F GMs) i s introd uceda t the elementl evel by means of t he twomos t commonlyused sche mes: (i ) n o dal b asedgr a da t ion, often via an a ux iliary (non -physical) te mperature- d e p endence, and (i i ) Gauss integr ation point bas e d gradat ion.These fo rmulation s are exte nsi vel y c omp a r ed by solving a n umbe rof para dig matic b oun dar y v alu eproblemsfor whic han al yt ica l sol u tions ca nbeob tai ned.T he nat ure o f th eno t abl e diffe r en c e s re ve al ed b y t he resu ltsi s i nvestig ated in d eta i l. W epr ovide a user subrout in e for thefi nit e elem e n t packag e ABAQUS to overcome th e limita tio ns of the most pop ula r appr oac h for i mpleme nting F GMs i n com m e rc ial s oftware. T h e us e ofre liab le, ele ment-based formula t ion s to define t hemate r i al pr o pe r tyva r iat i o n could be keyin fractur ea ss essment of FGM sand oth er non- homog e neous m aterials. ' address :'Dep a r tme nt of Engi neering, Universi t y ofC am bridg e,CB2 1P ZCam bridg e, UK' aut hor:- 'Emi li o Mart ínez- Pa ñeda' bi bliography: - 'library. bib' t itle: On the Fini teE lem ent Imple ment ation of

Tlidi,_M. Georgiou,_and P. Mandel, Phys._Rev. A_**48**,_4605 (1993);_M._Tlidi, P. Mandel,_and R. Lefever,_Phys. Rev. Lett. **73**,_640 (1994). V. J._Sánchez-Morcillo_and G. J. de Valcárcel, Quantum Semiclass. Opt. **8**, 919 (1996). A. H. Nayfeh, *Perturbation_methods*_(John Wiley_&_Sons,_New York, 2000) G. Kozyreff, S._J. Chapman, and M. Tlidi,_Phys. Rev._E** 68**, 015201(R) (2003); G. Kozyreff and M. Tlidi,_Phys._Rev. E **69**,_066202 (2004). C. Durniak *et al.*, Phys. Rev. E** 72**, 026607_(2005). M. G. Clerc, A. Petrossian, and S._Residori, Phys. Rev._E,_**71,**_015205 (R) (2005). --- abstract:_'We investigate the numerical implementation of_functionally graded properties in the context_of the finite element method. The macroscopic_variation of elastic properties inherent to_functionally graded materials (FGMs) is_introduced at_the element level by means_of the two_most commonly_used schemes: (i)_nodal based gradation, often via an_auxiliary (non-physical) temperature-dependence,_and (ii) Gauss integration point based_gradation._These formulations are_extensively_compared_by solving_a number of_paradigmatic_boundary value_problems_for which analytical solutions can be_obtained._The nature of the notable differences revealed_by the results is_investigated_in detail. We provide_a user subroutine for the_finite element package ABAQUS to overcome_the limitations_of the_most popular approach for implementing FGMs in commercial software. The use_of reliable, element-based formulations to define_the material property variation_could be_key_in fracture assessment_of_FGMs and_other non-homogeneous materials.' address: 'Department of Engineering, University_of Cambridge,_CB2 1PZ Cambridge, UK' author: - 'Emilio Martínez-Pañeda' bibliography: -_'library.bib' title: On the Finite_Element_Implementation of

approx \frac{\Lambda \rho ^3}{6M}\;\;\;\;\;\;\;\;\;\; \label{28}$$ The obtained metric does not change sign and has quite different behavior in the intervals which Schwarzschild-deSitter metric faces with problems. In contrast to Schwarzschild-deSitter metric that at origin gives $% g_{tt}=+\infty $ and $g_{rr}=0$, this new metric gives $g_{tt}=0$ and $% g_{rr}=+\infty $. thus the non-static metric is analytic everywhere except at $\rho =0$. At $\;\rho =2M-\frac{M^3\Lambda }3$ that is the familiar horizon of black holes the metric gives $$A(2M-\frac{M^3\Lambda }3)\approx \sqrt{\frac \Lambda 3}\times 2M \label{29}$$ On geodesic with $d\theta =d\varphi =0$, the relation between $\;dt$ and $dr$ at this point for a massless particle is $$dt=\frac{dr}{\sqrt{\frac 43\Lambda M^2}} \label{30}$$ Since $\Lambda $ is of the order $10^{-56}cm^{-2}$, then it takes less than $% \tau \equiv 10^5(\frac M{km})^{-1}$ years for a photon to escape from horizon region of black hole i.e. $\rho \approx 2M$ to outside of few $M.$ Since the metric is nonsingular for $r\neq 0$ and there is no horizon, the high potential regions are accessible. This huge amount of potential energy can be converted to thermal or other physical forms of energy and providing us with a new supermachinary for high energy astrophysics. This can make drastic change in our understanding of the AGN phenomenon. We should mention that the presented non-static metric and the Schwarzschild-deSitter metric are transformable to each other by the following coordinate transformation $$\rho =R(t)\;r\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;$$ $$T=t+\frac{\dot{R}}R\int^\rho \frac{A(\tilde{\rho})\tilde{\rho}d\tilde{\rho}}{% 1-\frac{2M}{\tilde{\rho}}-\frac \Lambda 3\tilde{\rho}^2} \label

approx \frac{\Lambda \rho ^3}{6M}\;\;\;\;\;\;\;\;\;\; \label{28}$$ The obtained metric does not change sign and take quite unlike behavior in the intervals which Schwarzschild - deSitter measured faces with problem. In contrast to Schwarzschild - deSitter metric function that at origin gives $% g_{tt}=+\infty $ and $ g_{rr}=0 $, this fresh metric gives $ g_{tt}=0 $ and $% g_{rr}=+\infty $. therefore the non - static metric is analytic everywhere except at $ \rho = 0$. At $ \;\rho = 2M-\frac{M^3\Lambda } 3 $ that is the familiar horizon of black fix the metric gives $ $ A(2M-\frac{M^3\Lambda } 3)\approx \sqrt{\frac \Lambda 3}\times 2 M \label{29}$$ On geodesic with $ d\theta = d\varphi = 0 $, the relation between $ \;dt$ and $ dr$ at this point for a massless particle is $ $ dt=\frac{dr}{\sqrt{\frac 43\Lambda M^2 } } \label{30}$$ Since $ \Lambda $ is of the ordering $ 10^{-56}cm^{-2}$, then it takes less than $% \tau \equiv 10 ^ 5(\frac M{km})^{-1}$ years for a photon to escape from horizon area of black hole i.e. $ \rho \approx 2M$ to outside of few $ M.$ Since the metric function is nonsingular for $ r\neq 0 $ and there is no horizon, the eminent potential regions are accessible. This huge amount of potential energy can be convert to thermal or other physical forms of energy and providing us with a newfangled supermachinary for high energy astrophysics. This can make drastic change in our understanding of the AGN phenomenon. We should mention that the presented non - static metric and the Schwarzschild - deSitter metric are convertible to each early by the following coordinate transformation $ $ \rho = R(t)\;r\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;$$ $ $ T = t+\frac{\dot{R}}R\int^\rho \frac{A(\tilde{\rho})\tilde{\rho}d\tilde{\rho}}{% 1-\frac{2M}{\tilde{\rho}}-\frac \Lambda 3\tilde{\rho}^2 } \label

appgox \frac{\Lambda \rho ^3}{6M}\;\;\;\;\;\;\;\;\;\; \lanel{28}$$ The obtained metric voes nof change sign and has quite differenv beyavioe in the intervals whizh Schwardschild-deWitttr metric faces wmfh problems. Ih concrest to Schwarzsghild-deSittar metric that ag lrigin gives $% g_{tt}=+\infty $ and $g_{rr}=0$, this new meyrlc gives $g_{tt}=0$ agd $% g_{gr}=+\ynftg $. thus the non-static metric is anzlytic tverywhere except st $\rho =0$. At $\;\rho =2M-\frac{M^3\Lambdw }3$ tjat is the familiag horizon od blwxk holes the metric ginzs $$A(2M-\frac{M^3\Lajbda }3)\approx \sqrt{\frac \Lambda 3}\timds 2M \label{29}$$ On ggosedhc with $d\thxta =d\vwrphi =0$, the rckation betweem $\;dt$ and $dr$ at thms piint for a massless perticle is $$dt=\frac{dr}{\sqtt{\frac 43\Lamtdc M^2}} \label{30}$$ Since $\Lambdq $ is ox tha orawr $10^{-56}zm^{-2}$, uhei if taked lxss than $% \tah \equiv 10^5(\frax M{km})^{-1}$ years for a pnoein to escape rrom hjryzon region of black hole i.e. $\rho \approx 2M$ fo outside of few $M.$ Sincw the metric is nonsijgular fow $r\neq 0$ and there is no horizon, the high potentian regmovs cvc aczwsdible. This huge amount of potential energy cag bt cpnverted to thcrmal or other phyxifak forms of enetgy and prkviding us with a jew supgrmachunary for hign energy astrophysics. This xan make drafric change in our bnderstandiny of tne AGM phenomenon. We should mzntion that the pgesented hun-static metric xnd tve Schwarzschild-deSitter mqtric are trausformabue tp each other by hhe following coordinate trandfmrmation $$\rhl =R(t)\;r\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;$$ $$T=t+\frac{\dot{R}}R\int^\rho \frac{A(\tilvx{\rho})\tilde{\rho}d\jilge{\rvo}}{% 1-\frac{2M}{\tnlde{\rhp}}-\frac \Lambda 3\eilde{\rho}^2} \labek

approx \frac{\Lambda \rho ^3}{6M}\;\;\;\;\;\;\;\;\;\; \label{28}$$ The obtained not sign and quite different behavior metric with problems. In to Schwarzschild-deSitter metric at origin gives $% g_{tt}=+\infty $ $g_{rr}=0$, this new metric gives $g_{tt}=0$ and $% g_{rr}=+\infty $. thus the non-static is analytic everywhere except at $\rho =0$. At $\;\rho =2M-\frac{M^3\Lambda }3$ that is familiar of holes metric gives $$A(2M-\frac{M^3\Lambda }3)\approx \sqrt{\frac \Lambda 3}\times 2M \label{29}$$ On geodesic with $d\theta =d\varphi =0$, the between $\;dt$ and $dr$ at this point for massless particle is $$dt=\frac{dr}{\sqrt{\frac M^2}} \label{30}$$ Since $\Lambda $ of order $10^{-56}cm^{-2}$, it less $% \tau \equiv M{km})^{-1}$ years for a photon to escape from horizon region of black hole i.e. $\rho \approx 2M$ outside of Since the is for 0$ and there horizon, the high potential regions are amount of potential energy can be converted to or other forms of energy and providing us a new supermachinary for high energy astrophysics. This make drastic change in our understanding of the AGN phenomenon. We should mention that the metric and the Schwarzschild-deSitter are transformable to other the coordinate $$\rho =R(t)\;r\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;$$ \frac{A(\tilde{\rho})\tilde{\rho}d\tilde{\rho}}{% 1-\frac{2M}{\tilde{\rho}}-\frac \Lambda 3\tilde{\rho}^2} \label

approx \frac{\Lambda \rho ^3}{6M}\;\;\;\;\;\;\;\;\;\; \label{28}$$ the obtaineD metrIc dOes NoT chaNge sIgn and has quite DIffeRent behavior in the interVals wHiCH SchWArZschiLd-deSitTEr METriC fAcEs wItH PrOblemS. In ContrasT to SchwarzSchIlD-deSitter metRIc That at origIn gIves $% g_{tt}=+\infty $ And $G_{rr}=0$, thiS nEw mETric gIveS $g_{tt}=0$ aNd $% g_{rr}=+\iNFty $. thuS the non-stAtIC metriC Is analyTIC eVeryWhere except at $\rho =0$. AT $\;\RhO =2m-\frac{M^3\Lambda }3$ thAt is thE fAMiLIAr hOriZon of black HoLes thE Metric gIVeS $$a(2m-\FraC{m^3\Lambda }3)\approx \Sqrt{\frac \LamBDa 3}\tImes 2M \lAbEl{29}$$ ON GeodesIc witH $d\THetA =d\varphi =0$, the RelaTion betweEn $\;dt$ anD $Dr$ at thiS Point foR a massLesS paRticLE iS $$dT=\frAc{DR}{\sqRT{\fRac 43\lAmbDa M^2}} \label{30}$$ siNcE $\LambDa $ is OF THE ordEr $10^{-56}cM^{-2}$, theN it taKes less than $% \taU \eqUiv 10^5(\fRAc M{Km})^{-1}$ yeaRs for A phoToN to esCape frOm horIzOn region of black Hole I.e. $\rho \apprOx 2M$ To OutSiDe of fEW $M.$ SincE thE meTric is nOnsinguLAr fOr $R\NEQ 0$ aNd there is no horizon, ThE HIgH potentiAl regiONs ArE AccessibLe. thiS hugE AMount Of poTEnTial enerGy can bE CoNvErted to ThErmal oR oTheR phYsicaL FormS of eneRgy and prOvidiNG us with a new supERmachinary for HIgH ENeRGy asTroPhysics. This Can mAKe drAstiC ChAngE In our UnderStANdINg of the AGN phenomenoN. WE shoulD mentIon that the preSented non-sTATIc metric And tHE SCHwarzschild-deSItter Metric are tRAnsformaBle to Each otheR by the folLOWing coorDinAte TraNsfORMaTion $$\rho =R(t)\;r\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;$$ $$T=t+\fRAC{\dot{r}}R\Int^\rho \fRac{a(\tilde{\rHo})\tIldE{\rhO}d\tIlDe{\rho}}{% 1-\frac{2m}{\tilde{\rhO}}-\fRaC \LAmBda 3\Tilde{\RHo}^2} \label

approx \frac{\Lambda \rho^3}{6M}\;\ ;\;\; \;\ ;\; \; \;\; \l abel{28}$$ Th e obt ained metric does notchang es igna nd hasquite d i ff e r ent b eh avi or in theint ervalswhich Schw arz sc hild-deSitte r m etric face s w ith problems . I n cont ra stt o Sch war zschi ld-deS i tter m etric tha ta t orig i n gives $ %g_{t t}=+\infty $ and$ g_ { rr}=0$, this n ew met ri c g i v es$g_ {tt}=0$ an d$% g_ { rr}=+\i n ft y $ . t h us the non-st atic metric isanalyt ic ev e rywher e exc ep t at $\rho =0$. At$\;\rho = 2M-\fr a c{M^3\L a mbda }3 $ that is th e fa m il ia r h or i zon of bl a ckholes th eme tricgive s $ $A(2 M-\ frac {M^3\ Lambda }3)\ap pro x \s q rt{ \frac \Lam bda3} \time s 2M \labe l{ 29}$$ On geode sicwith $d\t het a=d\ va rphi= 0$, th e r ela tion be tween $ \ ;dt $a n d $ dr$ at this pointfo r amassless parti c le i s $$dt=\ fr ac{ dr}{ \ s qrt{\ frac 43 \LambdaM^2}}\l ab el{30}$ $ Since $ \La mbd a $ i s ofthe or der $10^ {-56} c m^{-2}$, theni t takes lesst ha n $% \tau \e quiv 10^5(\ frac M{km })^{ - 1} $ y e ars f or aph o to n to escape from hor iz on reg ion o f black holei.e. $\rho \ a pprox 2M $ to ou t side of few $M .$ S ince the m e tric isnonsi ngular f or $r\neq 0 $ and th ere is no ho r i zo n, the high p o t enti al region s a re acce ssi ble . T his h uge amoun t of pot en ti al e ner gy ca n be conv er ted t o t herma l or ot her p hysi ca lf orm s of en e rg y andpr ov idin g u switha ne w su permach inary for hi g h en er gy astrop hysics. Thisca n make dra st icchange i n our un derstanding of the AGNp henomen on. Weshou ld mentio n t hat th e p r esente d non- stati cmet r i c and t he Sc hw arzschild- d e Sit ter m et ricare tra nsformable to each oth er by the fol low ingc o or din a te tra ns f orm a t ion $$\rho =R( t)\;r\;\;\ ;\ ; \; \;\;\;\;\; \ ;\; \; \;\;\;\ ;\;\;\; \;$$$$T=t+\ frac{\dot {R}}R\int ^\ rho\ f rac {A(\tilde{ \rho})\t ilde{\rho } d\til d e{ \rho} }{% 1-\fr ac {2M }{\ti lde{\r h o}} -\fra c \Lam bd a 3\ti lde{\ rh o}^2} \ label

approx \frac{\Lambda_\rho ^3}{6M}\;\;\;\;\;\;\;\;\;\;_ \label{28}$$ The obtained metric_does not_change_sign and_has_quite different behavior_in the intervals_which Schwarzschild-deSitter metric faces_with problems. In_contrast_to Schwarzschild-deSitter metric that at origin gives $% g_{tt}=+\infty $ and $g_{rr}=0$, this new metric_gives_$g_{tt}=0$ and_$% g_{rr}=+\infty_$._thus the non-static metric is_analytic everywhere except at $\rho_=0$. At_$\;\rho =2M-\frac{M^3\Lambda }3$ that is the familiar horizon_of_black holes the_metric gives $$A(2M-\frac{M^3\Lambda }3)\approx \sqrt{\frac \Lambda 3}\times 2M \label{29}$$ On_geodesic with $d\theta =d\varphi =0$, the_relation between $\;dt$_and_$dr$_at this point for_a massless particle is $$dt=\frac{dr}{\sqrt{\frac 43\Lambda M^2}}_ \label{30}$$ Since $\Lambda $ is of_the order $10^{-56}cm^{-2}$, then it takes less_than $% \tau \equiv 10^5(\frac M{km})^{-1}$ years_for a photon to escape_from horizon_region of black hole i.e._$\rho \approx 2M$_to outside_of few $M.$ Since_the metric is nonsingular for $r\neq_0$ and there_is no horizon, the high potential_regions_are accessible. This_huge_amount_of potential_energy can be_converted_to thermal_or_other physical forms of energy and_providing_us with a new supermachinary for high_energy astrophysics. This can_make_drastic change in our_understanding of the AGN phenomenon. We_should mention that the presented non-static_metric and_the Schwarzschild-deSitter_metric are transformable to each other by the following coordinate transformation $$\rho_=R(t)\;r\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;$$ $$T=t+\frac{\dot{R}}R\int^\rho \frac{A(\tilde{\rho})\tilde{\rho}d\tilde{\rho}}{% 1-\frac{2M}{\tilde{\rho}}-\frac \Lambda 3\tilde{\rho}^2} \label

consider ${{\mathcal C}}$ as a DG ${{\mathcal C}}^0$-module then we obtain a DG ${{\mathcal C}}\otimes \hat{S}$-module ${{\mathcal C}}\otimes _{\tau _{{{\mathcal C}}}}B{{\mathcal C}}$, which is also quasi-isomorphic to $k$. Deformations and the bar complex -------------------------------- Let ${{\mathcal C}}$ be a DG algebra and ${{\mathcal R}}$ a finite dimensional DG algebra. Then the DG algebras ${{\mathcal C}}\otimes {{\mathcal R}}$ and $\Hom _k({{\mathcal R}}^* ,{{\mathcal C}})$ are naturally isomorphic. The isomorphism $\theta :{{\mathcal C}}\otimes {{\mathcal R}}\to \Hom_k({{\mathcal R}}^*,{{\mathcal C}})$ is given by the formula $$\theta (c\otimes r)(f)= (-1)^{\bar{f}\bar{r}}cf(r)$$ for $f\in {{\mathcal R}}^*$, $c\in {{\mathcal C}}$, $r\in {{\mathcal R}}$. Assume that in the above lemma the DG algebra ${{\mathcal R}}$ is artinian with the maximal ideal $m\subset {{\mathcal R}}$. Note that under the isomorphism of this lemma the ideal ${{\mathcal C}}\otimes m$ is mapped to the ideal $\Hom _k(m^*, {{\mathcal C}})$. Thus in particular there is a natural bijection between the set $MC({{\mathcal C}}\otimes m)$ and the collection of twisting cochains in $\Hom _k(m^*,{{\mathcal C}})$. Assume in the above remark that the DG algebra ${{\mathcal C}}$ is augmented and that ${{\mathcal R}}\in {\operatorname{dgart}}_-$ (hence ${{\mathcal R}}^*$ is concentrated in nonnegative degrees). Then elements of $MC({{\mathcal C}}\otimes m)$ are in bijection with the set of admissible twisting cochains in $\Hom _k({{\mathcal R}}^*,{{\mathcal C}})$. ### Let ${{\mathcal C}}$ be an augmented DG algebra, ${{\mathcal R}}$ an artinian DG algebra with the maximal ideal $m$. Let $\tau \in \Hom _k({{\mathcal R}}^*,{{\mathcal C}})$ be an admissible twisting cochain and ${{\mathcal R}}^*\otimes _{\tau}{{\mathcal C}}$ be the corresponding DG $({{\mathcal C}}\otimes {{\mathcal R}})^0$-module. Let $\alpha \in MC({{\mathcal C}}\otimes m)$ be the element corresponding to $\tau$

consider $ { { \mathcal C}}$ as a DG $ { { \mathcal C}}^0$-module then we obtain a DG $ { { \mathcal C}}\otimes \hat{S}$-module $ { { \mathcal C}}\otimes _ { \tau _ { { { \mathcal C}}}}B{{\mathcal C}}$, which is also quasi - isomorphic to $ k$. contortion and the prevention complex -------------------------------- Let $ { { \mathcal C}}$ be a DG algebra and $ { { \mathcal R}}$ a finite dimensional DG algebra. Then the DG algebra $ { { \mathcal C}}\otimes { { \mathcal R}}$ and $ \Hom _ k({{\mathcal R}}^ * , { { \mathcal C}})$ are naturally isomorphic. The isomorphism $ \theta: { { \mathcal C}}\otimes { { \mathcal R}}\to \Hom_k({{\mathcal R}}^*,{{\mathcal C}})$ is given by the formula $ $ \theta (c\otimes r)(f)= (-1)^{\bar{f}\bar{r}}cf(r)$$ for $ f\in { { \mathcal R}}^*$, $ c\in { { \mathcal C}}$, $ r\in { { \mathcal R}}$. Assume that in the above lemma the DG algebra $ { { \mathcal R}}$ is artinian with the maximal ideal $ m\subset { { \mathcal R}}$. Note that under the isomorphism of this lemma the ideal $ { { \mathcal C}}\otimes m$ is map to the ideal $ \Hom _ k(m^ *, { { \mathcal C}})$. Thus in particular there embody a natural bijection between the set $ MC({{\mathcal C}}\otimes m)$ and the solicitation of twisting cochains in $ \Hom _ k(m^*,{{\mathcal C}})$. Assume in the above remark that the DG algebra $ { { \mathcal C}}$ is augmented and that $ { { \mathcal R}}\in { \operatorname{dgart}}_-$ (hence $ { { \mathcal R}}^*$ is concentrated in nonnegative degree). Then elements of $ MC({{\mathcal C}}\otimes m)$ are in bijection with the set of admissible twisting cochains in $ \Hom _ k({{\mathcal R}}^*,{{\mathcal C}})$. # # # Let $ { { \mathcal C}}$ be an augmented DG algebra, $ { { \mathcal R}}$ an artinian decigram algebra with the maximal ideal $ m$. Let $ \tau \in \Hom _ k({{\mathcal R}}^*,{{\mathcal C}})$ be an admissible twisting cochain and $ { { \mathcal R}}^*\otimes _ { \tau}{{\mathcal C}}$ be the equate DG $ ({ { \mathcal C}}\otimes { { \mathcal R}})^0$-module. Let $ \alpha \in MC({{\mathcal C}}\otimes m)$ be the element match to $ \tau$

cojsider ${{\mathcal C}}$ as a DG ${{\mathcal C}}^0$-modulg rhen wx obtaih a DG ${{\mxthcal C}}\otimes \hat{S}$-module ${{\matical C}}\otines _{\tau _{{{\mathcal C}}}}B{{\mathzal C}}$, whibh is alsi quesi-isomorphic to $k$. Deformations wnd chx bar complex -------------------------------- Lej ${{\mathcal C}}$ te a DG algebrd xnb ${{\mathcal R}}$ a finite dimensional DG wlgebra. Tjen the DG alggbras ${{\iathdal C}}\otimes {{\mathcal R}}$ and $\Hom _k({{\matgcal R}}^* ,{{\methcal C}})$ are nayurally isomorphic. The isomlrphlsm $\theta :{{\mathcal F}}\otimes {{\matycal E}}\to \Hom_k({{\mathcxl R}}^*,{{\mathcal C}})$ is given by the formula $$\theta (c\otimes r)(w)= (-1)^{\bar{f}\yar{r}}cf(r)$$ for $f\un {{\ldthcal R}}^*$, $c\ii {{\mathbal C}}$, $r\in {{\mathgsl R}}$. Asvume thst in the abovc lemka rhe DG algebra ${{\mathcan R}}$ is artinian wijh the maxhmcl ideal $m\subset {{\mathxao R}}$. Nmte dhat ynddr uhe iaomorpjisj of this memma the ieeal ${{\mathcal C}}\otimex i$ is mapped to the iqewl $\Hom _k(m^*, {{\mathcal C}})$. Thus in particular uhere is a natural bijection between the set $MC({{\majhcal C}}\otiies m)$ and the collection of twisting cochains in $\Vom _k(j^*,{{\oatkgwu C}})$. Wssume in the above remark that the DG algebrw ${{\mstmcal C}}$ is augmenbed and that ${{\mathcsl R}}\og {\operatorname{agart}}_-$ (ksnde ${{\mathcal R}}^*$ is cojcentrajed in nonnegatyve cegrees). Then elements of $MC({{\nathcal C}}\otiiws m)$ are in bijectnon with the set of acmissible twisting cochcins ih $\Hom _k({{\mathfal R}}^*,{{\mathdxl C}})$. ### Let ${{\mathcal C}}$ ne dn augmented DG algebra, ${{\maehcal R}}$ ai artnnian DG alggbra wieh the maxlmal lgeal $m$. Let $\tau \in \Jom _k({{\mcthcan R}}^*,{{\mathcal C}})$ be an admissible twisting coriain and ${{\mathvan R}}^*\mtimes _{\tcu}{{\mathgal C}}$ be the cowresponding DG $({{\mathcaj C}}\otkmes {{\mathcam R}})^0$-modune. Let $\alpha \in MC({{\mathcal F}}\otimes m)$ be the eleient coreespondkvg to $\tau$

consider ${{\mathcal C}}$ as a DG ${{\mathcal we a DG C}}\otimes \hat{S}$-module ${{\mathcal which also quasi-isomorphic to Deformations and the complex -------------------------------- Let ${{\mathcal C}}$ be DG algebra and ${{\mathcal R}}$ a finite dimensional DG algebra. Then the DG ${{\mathcal C}}\otimes {{\mathcal R}}$ and $\Hom _k({{\mathcal R}}^* ,{{\mathcal C}})$ are naturally isomorphic. isomorphism :{{\mathcal {{\mathcal \Hom_k({{\mathcal R}}^*,{{\mathcal C}})$ is given by the formula $$\theta (c\otimes r)(f)= (-1)^{\bar{f}\bar{r}}cf(r)$$ for $f\in {{\mathcal R}}^*$, {{\mathcal C}}$, $r\in {{\mathcal R}}$. Assume that in above lemma the DG ${{\mathcal R}}$ is artinian with maximal $m\subset {{\mathcal Note under isomorphism of this the ideal ${{\mathcal C}}\otimes m$ is mapped to the ideal $\Hom _k(m^*, {{\mathcal C}})$. Thus in particular is a between the $MC({{\mathcal m)$ the collection of in $\Hom _k(m^*,{{\mathcal C}})$. Assume in that the DG algebra ${{\mathcal C}}$ is augmented that ${{\mathcal {\operatorname{dgart}}_-$ (hence ${{\mathcal R}}^*$ is concentrated nonnegative degrees). Then elements of $MC({{\mathcal C}}\otimes m)$ in bijection with the set of admissible twisting cochains in $\Hom _k({{\mathcal R}}^*,{{\mathcal C}})$. ### C}}$ be an augmented algebra, ${{\mathcal R}}$ artinian algebra the ideal $m$. $\tau \in \Hom _k({{\mathcal R}}^*,{{\mathcal C}})$ be an admissible twisting cochain ${{\mathcal R}}^*\otimes _{\tau}{{\mathcal C}}$ be the corresponding DG $({{\mathcal C}}\otimes Let \in MC({{\mathcal C}}\otimes be the element corresponding $\tau$

consider ${{\mathcal C}}$ as a DG ${{\mathCal C}}^0$-module Then wE obTaiN a dG ${{\maThcaL C}}\otimes \hat{S}$-moDUle ${{\mAthcal C}}\otimes _{\tau _{{{\mathcaL C}}}}B{{\maThCAl C}}$, wHIcH is alSo quasi-ISoMORphIc To $K$. DeFoRMaTions And The bar cOmplex -------------------------------- Let ${{\mAthCaL C}}$ be a DG algebRA aNd ${{\mathcal R}}$ A fiNite dimensioNal dG algeBrA. ThEN the Dg alGebraS ${{\mathcAL C}}\otimEs {{\mathcal r}}$ aND $\Hom _k({{\mAThcal R}}^* ,{{\mATHcAl C}})$ aRe naturally isomorPHiC. the isomorphism $\Theta :{{\mAtHCaL c}}\OtiMes {{\Mathcal R}}\to \hoM_k({{\matHCal R}}^*,{{\matHCaL c}})$ IS giVEn by the formulA $$\theta (c\otimES r)(f)= (-1)^{\Bar{f}\baR{r}}Cf(r)$$ FOr $f\in {{\mAthcaL R}}^*$, $C\In {{\mAthcal C}}$, $r\in {{\mAthcAl R}}$. Assume That in THe above LEmma the dG algeBra ${{\MatHcal r}}$ Is ArTinIaN WitH ThE maXImaL ideal $m\sUbSeT {{\mathCal R}}$. nOTE That UndEr thE isomOrphism of this LemMa thE IdeAl ${{\matHcal C}}\OtimEs M$ is maPped to The idEaL $\Hom _k(m^*, {{\mathcal C}})$. THus iN particulAr tHeRe iS a NaturAL bijecTioN beTween thE set $MC({{\mAThcAl c}}\OTImEs m)$ and the collectioN oF TWiSting cocHains iN $\hoM _k(M^*,{{\Mathcal C}})$. asSumE in tHE Above RemaRK tHat the DG AlgebrA ${{\MaThCal C}}$ is aUgMented AnD thAt ${{\mAthcaL r}}\in {\oPeratoRname{dgaRt}}_-$ (henCE ${{\mathcal R}}^*$ is conCEntrated in nonNEgATIvE DegrEes). then elementS of $Mc({{\MathCal C}}\OTiMes M)$ Are in BijecTiON wITh the set of admissiblE tWistinG cochAins in $\Hom _k({{\matHcal R}}^*,{{\mathcAL c}})$. ### let ${{\mathcAl C}}$ bE An AUgmented DG algeBra, ${{\maThcal R}}$ an arTInian DG aLgebrA with the Maximal idEAL $m$. Let $\tau \In \HOm _k({{\MatHcaL r}}^*,{{\MaThcal C}})$ be an admISSiblE tWisting CocHain and ${{\MatHcaL R}}^*\oTimEs _{\Tau}{{\mathcaL C}}$ be the cOrReSpOnDinG DG $({{\maTHcal C}}\otiMeS {{\maThCal r}})^0$-moduLE. Let $\alPha \in mC({{\maThCaL c}}\otImes m)$ be THe ELEmenT cOrRespOndInG to $\taU$

consider ${{\mathcal C}}$ as a DG $ {{\ma thc alC} }^0$ -mod ule then we ob t aina DG ${{\mathcal C}}\o times \ h at{S } $- modul e ${{\m a th c a l C }} \o tim es _{ \tau_{{ {\mathc al C}}}}B{ {\m at hcal C}}$, w h ic h is alsoqua si-isomorphi c t o $k$. Def o rmati ons andthe ba r compl ex ------ -- - ------ - ------- - - -- ---- - Let ${{\mathca l C } }$ be a DG alg ebra a nd ${ { \ mat hca l R}}$ a f in ite d i mension a lD G alg e bra. Then the DG algebra s ${ {\math ca l C } }\otim es {{ \m a thc al R}}$ and $\H om _k({{\ mathca l R}}^*, {{\math cal C} })$ ar e na t ur al lyis o mor p hi c.The isomorp hi sm $\th eta: { { \ math cal C}} \otim es {{\mathcal R} }\to \Ho m_k({ {\mat hcal R }}^*, {{\mat hcalC} })$ is given by the formula$$\ th eta ( c\oti m es r)( f)= (- 1)^{\ba r{f}\ba r {r} }c f ( r )$ $ for $f\in {{\mat hc a l R }}^*$, $ c\in { { \m at h cal C}}$ ,$r\ in { { \ mathc al R } }$ . Assum e that in t he abov elemmath e D G a lgebr a ${{ \mathc al R}}$is ar t inian with the maximal ideal $m \ s ub s et { {\m athcal R}}$ . No t e th at u n de r t h e iso morph is m o f this lemma the ide al ${{\m athca l C}}\otimesm$ is mapp e d to the i deal $\ H om _k(m^*, {{\ mathc al C}})$.T hus in p artic ular the re is a n a t ural bij ect ion be twe e n t he set $MC({{ \ m athc al C}}\ot ime s m)$ a ndthe co lle ct ion of tw isting c oc ha in sin$\Hom _k(m^*,{ {\ mat hc alC}})$ . Assu me in the a bo v e r emark t h at t he D Gal gebr a $ {{ \math calC }}$ is aug mented an d t h at $ {{ \m athcalR}}\in {\oper at orname{dga rt }}_ -$ (he n c e ${{\ma thcal R}}^*$ is concent r ated in no nnega tive degrees) . T hen el eme n ts of$MC({{ \math ca l C } } \otim e s m )$ar e in bijec t i onwithth e se t of ad missible twistingc och ains in $\Hom _k ({{\ m a th cal R} } ^*, {{ \ mat h c al C}})$. ### Let ${{\ ma t hc al C}}$ be anau gmented DG alg ebra, ${{\mat hcal R}}$ an artin ia n DG a lge bra with t he maxim al ideal$ m$. L e t$\tau \i n \Hom _ k({ {\mat hcal R } }^* ,{{\m athcal C }})$ b e anad missible twisting cochain and $ {{\mat hcalR}} ^*\otimes _{ \ tau }{{\mathc al C }}$ be the co rre spond ing DG $( {{\m a th cal C}}\o time s {{\mathc a lR}} ) ^ 0$ -module. L e t $\a lpha\in MC({{\ math cal C}}\otimes m) $ be the elemen t co r r esp ond i ng t o$\tau$

consider_${{\mathcal C}}$_as a DG ${{\mathcal_C}}^0$-module then_we_obtain a_DG_${{\mathcal C}}\otimes \hat{S}$-module ${{\mathcal_C}}\otimes _{\tau _{{{\mathcal_C}}}}B{{\mathcal C}}$, which is_also quasi-isomorphic to_$k$. Deformations_and the bar complex -------------------------------- Let ${{\mathcal C}}$ be a DG algebra and ${{\mathcal R}}$ a_finite_dimensional DG_algebra._Then_the DG algebras ${{\mathcal C}}\otimes_{{\mathcal R}}$ and $\Hom _k({{\mathcal_R}}^* ,{{\mathcal C}})$_are naturally isomorphic. The isomorphism $\theta :{{\mathcal C}}\otimes {{\mathcal_R}}\to \Hom_k({{\mathcal_R}}^*,{{\mathcal C}})$ is_given by the formula $$\theta (c\otimes r)(f)= (-1)^{\bar{f}\bar{r}}cf(r)$$ for $f\in_{{\mathcal R}}^*$, $c\in {{\mathcal C}}$, $r\in {{\mathcal_R}}$. Assume that in_the_above_lemma the DG algebra_${{\mathcal R}}$ is artinian with the_maximal ideal $m\subset {{\mathcal R}}$. Note_that under the isomorphism of this lemma_the ideal ${{\mathcal C}}\otimes m$ is_mapped to the ideal $\Hom__k(m^*, {{\mathcal_C}})$. Thus in particular there_is a natural_bijection between_the set $MC({{\mathcal_C}}\otimes m)$ and the collection of_twisting cochains in_$\Hom _k(m^*,{{\mathcal C}})$. Assume in the above_remark_that the DG_algebra_${{\mathcal_C}}$ is_augmented and that_${{\mathcal_R}}\in {\operatorname{dgart}}_-$ (hence_${{\mathcal_R}}^*$ is concentrated in nonnegative degrees)._Then_elements of $MC({{\mathcal C}}\otimes m)$ are in_bijection with the set_of_admissible twisting cochains in_$\Hom _k({{\mathcal R}}^*,{{\mathcal C}})$. ### Let_${{\mathcal C}}$ be an augmented DG_algebra, ${{\mathcal_R}}$ an_artinian DG algebra with the maximal ideal $m$. Let $\tau \in_\Hom _k({{\mathcal R}}^*,{{\mathcal C}})$ be an admissible_twisting cochain and ${{\mathcal_R}}^*\otimes _{\tau}{{\mathcal_C}}$_be the corresponding_DG_$({{\mathcal C}}\otimes {{\mathcal_R}})^0$-module. Let $\alpha \in MC({{\mathcal C}}\otimes m)$ be_the element_corresponding to $\tau$

hyperbolic. (The existence of such $D$ was shown by Zaidenberg in [@zai89].) Let $S$ be a generic small deformation of $2l D$. Then the double cover of ${\mathbb{P}}^2$ branched along $S$ is Brody hyperbolic. Apply Theorem \[mainthm2\] to $(X,L,H,d,m):=({\mathbb{P}}^2, {\mathcal{O}}(k), D, 2, 2l)$. \[k3\] Let $(X_0,L_0)$ be a primitively polarized K3 surface of degree $2l$ for $l\in{\mathbb{Z}}_{>0}$. For any $m\geq 4$, denote by $M_0:=L_0^m$. Pick a general member $H\in |L_0|$, then $H$ is smooth and $g(H)\geq 2$. Let $S_0$ be a generic small deformation of $mH$ that is smooth. Then Theorem \[complement\] implies that $S_0$ and $X_0\setminus S_0$ are both Brody hyperbolic. There exists a deformation $({\mathcal{X}},{\mathcal{M}})$ of $(X_0, M_0)$ over $\Delta$ such that $(X_t,M_t)$ is a primitively polarized K3 surface of degree $2lm^2$ for $t\in\Delta\setminus\{0\}$. It is clear that $h^0(X_t,M_t)$ does not depend on the choice of $t$ in $\Delta$. Hence Grauert’s theorem implies that $\pi_*{\mathcal{M}}$ is a locally free sheaf on $\Delta$, where $\pi:{\mathcal{X}}\to\Delta$ is the projection map. In other words, $\{H^0(X_t,M_t)\}_{t\in\Delta}$ forms a holomorphic vector bundle over $\Delta$. Now we may deform $S_0$ to a family of divisors $S_t\in |M_t|$ for $|t|$ sufficiently small. By choosing a generic deformation, we may assume that $S_t$ is smooth for $|t|$ sufficiently small. Hence Theorem \[def-comp\] implies that both $S_t$ and $X_t\setminus S_t$ are Brody hyperbolic for $|t|$ sufficiently small

hyperbolic. (The existence of such $ D$ was shown by Zaidenberg in [ @zai89 ] .) Let $ S$ be a generic small distortion of $ 2l D$. Then the bivalent cover of $ { \mathbb{P}}^2 $ branched along $ S$ is Brody hyperbolic. lend oneself Theorem \[mainthm2\ ] to $ (X, L, H, d, m):=({\mathbb{P}}^2, { \mathcal{O}}(k), D, 2, 2l)$. \[k3\ ] get $ (X_0,L_0)$ be a primitively polarized K3 airfoil of academic degree $ 2l$ for $ l\in{\mathbb{Z}}_{>0}$. For any $ m\geq 4 $, denote by $ M_0:=L_0^m$. Pick a general member $ H\in |L_0|$, then $ H$ is smooth and $ g(H)\geq 2$. lease $ S_0 $ be a generic small deformation of $ mH$ that is legato. Then Theorem \[complement\ ] implies that $ S_0 $ and $ X_0\setminus S_0 $ are both Brody hyperbolic. There exists a deformation $ ({ \mathcal{X}},{\mathcal{M}})$ of $ (X_0, M_0)$ over $ \Delta$ such that $ (X_t, M_t)$ is a primitively polarized K3 surface of academic degree $ 2lm^2 $ for $ t\in\Delta\setminus\{0\}$. It is clear that $ h^0(X_t, M_t)$ does not depend on the choice of $ t$ in $ \Delta$. therefore Grauert ’s theorem implies that $ \pi_*{\mathcal{M}}$ is a locally free sheaf on $ \Delta$, where $ \pi:{\mathcal{X}}\to\Delta$ is the projection map. In early words, $ \{H^0(X_t, M_t)\}_{t\in\Delta}$ shape a holomorphic vector bundle over $ \Delta$. Now we may deform $ S_0 $ to a syndicate of divisors $ S_t\in |M_t|$ for $ |t|$ sufficiently small. By choosing a generic deformation, we may assume that $ S_t$ is smooth for $ |t|$ sufficiently small. Hence Theorem \[def - comp\ ] implies that both $ S_t$ and $ X_t\setminus S_t$ are Brody hyperbolic for $ |t|$ sufficiently small

hyoerbolic. (The existence on such $D$ was shocb by Zeidenbedg in [@zak89].) Let $S$ be a generic small dxfornatiob of $2l D$. Then the doubue cover lf ${\mathbv{P}}^2$ bcanched along $S$ ma Brody hypergllic. Cp'ly Theorem \[maikthm2\] to $(X,L,H,g,m):=({\mathbb{P}}^2, {\mathwau{O}}(n), D, 2, 2l)$. \[k3\] Let $(X_0,L_0)$ be a primitively polwrized L3 durface of degtee $2l$ sor $m\pn{\nathbb{Z}}_{>0}$. For any $m\geq 4$, denote gy $M_0:=L_0^m$. Kick a general memner $H\in |L_0|$, then $H$ is smooth and $g(H)\geq 2$. Let $S_0$ be a generic smqll qwformation ow $mH$ that pv smooth. Tgen Theorem \[complement\] implies ghat $X_0$ and $X_0\setnibus V_0$ are both Urody ryperbolic. Thcge exisds a degormation $({\mathgal{X}},{\methcql{M}})$ of $(X_0, M_0)$ over $\Delta$ such that $(X_t,M_t)$ is a primithvzly polarized K3 surfaxe of dggree $2lm^2$ wir $g\in\Sektz\setmijus\{0\}$. It is clezr that $h^0(X_t,N_t)$ does not depend pn nne choice of $t$ in $\Qejta$. Hence Grauert’s theorem implies that $\pi_*{\jathcal{M}}$ is a locally feee sheaf on $\Delta$, whgre $\pi:{\mathsal{X}}\to\Delta$ is the projection map. In other words, $\{V^0(X_t,M_t)\}_{v\iv\Deotw}$ dogms a holomorphic vector bundle over $\Delta$. Nor wt msy deform $S_0$ to a family og fifysors $S_t\in |M_t|$ for $|t|$ surficiently small. Bj choosyng a generic qefotmation, we may assume that $W_t$ is smooth dor $|t|$ sufficiently small. Hencz Theotem \[deg-comp\] implies that both $S_t$ znd $X_t\setmijus S_t$ ars Brody hyperboliz fpr $|t|$ sufficiently small

hyperbolic. (The existence of such $D$ was Zaidenberg [@zai89].) Let be a generic Then double cover of branched along $S$ Brody hyperbolic. Apply Theorem \[mainthm2\] to {\mathcal{O}}(k), D, 2, 2l)$. \[k3\] Let $(X_0,L_0)$ be a primitively polarized K3 surface degree $2l$ for $l\in{\mathbb{Z}}_{>0}$. For any $m\geq 4$, denote by $M_0:=L_0^m$. Pick a member |L_0|$, $H$ smooth and $g(H)\geq 2$. Let $S_0$ be a generic small deformation of $mH$ that is smooth. Theorem \[complement\] implies that $S_0$ and $X_0\setminus S_0$ both Brody hyperbolic. There a deformation $({\mathcal{X}},{\mathcal{M}})$ of $(X_0, over such that is primitively K3 surface of $2lm^2$ for $t\in\Delta\setminus\{0\}$. It is clear that $h^0(X_t,M_t)$ does not depend on the choice of $t$ in Hence Grauert’s that $\pi_*{\mathcal{M}}$ a free on $\Delta$, where the projection map. In other words, holomorphic vector bundle over $\Delta$. Now we may $S_0$ to family of divisors $S_t\in |M_t|$ for sufficiently small. By choosing a generic deformation, we assume that $S_t$ is smooth for $|t|$ sufficiently small. Hence Theorem \[def-comp\] implies that both $X_t\setminus S_t$ are Brody for $|t|$ sufficiently

hyperbolic. (The existence of sUch $D$ was shoWn by ZAidEnbErG in [@zAi89].) LeT $S$ be a generic smALl deFormation of $2l D$. Then the doUble cOvER of ${\mAThBb{P}}^2$ brAnched aLOnG $s$ Is BRoDy HypErBOlIc. AppLy THeorem \[mAinthm2\] to $(X,L,h,d,m):=({\MaThbb{P}}^2, {\mathcal{o}}(K), D, 2, 2L)$. \[k3\] Let $(X_0,L_0)$ be a PriMitively polaRizEd K3 surFaCe oF DegreE $2l$ fOr $l\in{\Mathbb{z}}_{>0}$. for any $M\geq 4$, denotE bY $m_0:=L_0^m$. PicK A generaL MEmBer $H\In |L_0|$, then $H$ is smooth aND $g(h)\Geq 2$. Let $S_0$ be a geneRic smaLl DEfORMatIon Of $mH$ that is SmOoth. THEn TheorEM \[cOMPLemENt\] implies that $s_0$ and $X_0\setminUS S_0$ aRe both brOdy HYperboLic. ThErE ExiSts a deformaTion $({\Mathcal{X}},{\mAthcal{m}})$ Of $(X_0, M_0)$ oveR $\delta$ suCh that $(x_t,M_T)$ is A priMItIvEly PoLAriZEd k3 suRFacE of degreE $2lM^2$ fOr $t\in\deltA\SETMinuS\{0\}$. It Is clEar thAt $h^0(X_t,M_t)$ does noT dePend ON thE choiCe of $t$ In $\DeLtA$. HencE GraueRt’s thEoRem implies that $\pI_*{\matHcal{M}}$ is a lOcaLlY frEe Sheaf ON $\Delta$, WheRe $\pI:{\mathcaL{X}}\to\DelTA$ is ThE PROjEction map. In other woRdS, $\{h^0(x_t,m_t)\}_{t\in\DelTa}$ formS A hOlOMorphic vEcTor BundLE Over $\DElta$. nOw We may defOrm $S_0$ to A FaMiLy of divIsOrs $S_t\iN |M_T|$ foR $|t|$ sUfficIEntlY small. by choosiNg a geNEric deformatioN, We may assume thAT $S_T$ IS sMOoth For $|T|$ sufficientLy smALl. HeNce THEoRem \[DEf-comP\] implIeS ThAT both $S_t$ and $X_t\setminuS S_T$ are BrOdy hyPerbolic for $|t|$ sUfficientlY SMAll

hyperbolic. (The existenc e of such$D$ w assho wn byZaid enberg in [@za i 89]. ) Let $S$ be a generic smal ld efor m at ion o f $2l D $ .T h enth edou bl e c overof${\math bb{P}}^2$bra nc hed along $S $ i s Brody hy per bolic. Appl y T heorem \ [ma i nthm2 \]to $( X,L,H, d ,m):=( {\mathbb{ P} } ^2, {\ m athcal{ O } }( k),D, 2, 2l)$. \[k3 \ ]L et $(X_0,L_0)$ be apr i mi t i vel y p olarized K 3surfa c e of de g re e $ 2l$ for $l\in{\ma thbb{Z}}_{> 0 }$. For a ny $m \ geq 4$ , den ot e by $M_0:=L_0^ m$.Pick a ge neralm ember $ H \in |L_ 0|$, t hen $H $ is sm oo than d $g ( H) \ge q 2$ . Let $S _0 $be agene r i c smal l d efor matio n of $mH$ tha t i s sm o oth . The n The orem \ [comp lement \] im pl ies that $S_0$and$X_0\setm inu sS_0 $are b o th Bro dyhyp erbolic . Ther e ex is t s adeformation $({\ma th c a l{ X}},{\ma thcal{ M }} )$ of $(X_0 ,M_0 )$ o v e r $\D elta $ s uch that $(X_t , M_ t) $ is apr imitiv el y p ola rized K3 s urface of degr ee $2 l m^2$ for $t\in \ Delta\setminu s \{ 0 \ }$ . Itisclear that$h^0 ( X_t, M_t) $ d oes not d epend o n t h e choice of $t$ in$\ Delta$ . Hen ce Grauert’stheorem im p l i es that$\pi _ *{ \ mathcal{M}}$ i s a l ocally fre e sheaf o n $\D elta$, w here $\pi : { \mathcal {X} }\t o\D elt a $ i s the project i o n ma p. In oth erwords,$\{ H^0 (X_ t,M _t )\}_{t\in \Delta}$ f or ms a ho lomor p hic vect or bu nd leover$ \Delta $. No w we m ay def orm $S_ 0 $t o a f am il y of di vi sors$S_t \ in|M_t|$for $|t|$ su f fici en tl y small . By choosing a generic d ef orm ation, w e may as sume that $S_t$ is smoo t h for $ |t| $ suf fici ently sma ll. Hence Th e orem \ [def-c omp\] i mpl i e s tha t bo th$S _t$ and $X _ t \se tminu sS_t$ are Br ody hyperbolic for $|t |$ sufficient lysmal l

hyperbolic._(The existence_of such $D$ was_shown by_Zaidenberg_in [@zai89].)_Let_$S$ be a_generic small deformation_of $2l D$. Then_the double cover_of_${\mathbb{P}}^2$ branched along $S$ is Brody hyperbolic. Apply Theorem \[mainthm2\] to $(X,L,H,d,m):=({\mathbb{P}}^2, {\mathcal{O}}(k), D, 2,_2l)$. \[k3\]_Let $(X_0,L_0)$_be_a_primitively polarized K3 surface of_degree $2l$ for $l\in{\mathbb{Z}}_{>0}$. For_any $m\geq_4$, denote by $M_0:=L_0^m$. Pick a general member_$H\in_|L_0|$, then $H$_is smooth and $g(H)\geq 2$. Let $S_0$ be a_generic small deformation of $mH$ that_is smooth. Then_Theorem_\[complement\]_implies that $S_0$ and_$X_0\setminus S_0$ are both Brody hyperbolic. There_exists a deformation $({\mathcal{X}},{\mathcal{M}})$ of $(X_0,_M_0)$ over $\Delta$ such that $(X_t,M_t)$ is_a primitively polarized K3 surface of_degree $2lm^2$ for $t\in\Delta\setminus\{0\}$. It_is clear_that $h^0(X_t,M_t)$ does not depend_on the choice_of $t$_in $\Delta$. Hence_Grauert’s theorem implies that $\pi_*{\mathcal{M}}$ is_a locally free_sheaf on $\Delta$, where $\pi:{\mathcal{X}}\to\Delta$ is_the_projection map. In_other_words,_$\{H^0(X_t,M_t)\}_{t\in\Delta}$ forms_a holomorphic vector_bundle_over $\Delta$._Now_we may deform $S_0$ to a_family_of divisors $S_t\in |M_t|$ for $|t|$ sufficiently_small. By choosing a_generic_deformation, we may assume_that $S_t$ is smooth for_$|t|$ sufficiently small. Hence Theorem \[def-comp\]_implies that_both $S_t$_and $X_t\setminus S_t$ are Brody hyperbolic for $|t|$ sufficiently small

Finally, let $\delta_i\in \partial(\zeta^{{\cal E}})$ for $i=0,..,j$. We have $$\begin{aligned} \nonumber &{}&\lim\limits_{n\to \infty}{\mathbb{P}}_{\delta_0} (Y_i=\delta_i, i=1,..,j \, | \, \zeta>n)\\ \nonumber &{}& = \lim\limits_{n\to \infty} {\mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,..,j) \frac{{\mathbb{P}}_{\delta_j}(\zeta>n-j)} {{\mathbb{P}}_{\delta_0}(\zeta>n)}\\ \label{rats} &{}&={\mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,..,j) \frac{\varphi_{\delta_j}}{\varphi_{\delta_0}}\eta^{-j}\\ \nonumber &{}& =\prod_{l=0}^{j-1} \left(\eta^{-1}P_{\delta_l, \delta_{l+1}}\, \frac{\varphi_{\delta_{l+1}}}{\varphi_{\delta_l}}\right).\end{aligned}$$ In (\[rats\]) we used $\lim\limits_{n\to \infty}{\mathbb{P}}(\zeta>n-j)/{\mathbb{P}}(\zeta>n)=\eta^{-j}$, which is a consequence of (\[50e\]). Then the result follows. Let $\nu=(\nu_\delta: \delta\in \partial({{\cal T}})^*)$ be a probability measure on $\partial({{\cal T}})^*$. If necessary, $\nu$ will be identified with its extension on $\partial({{\cal T}})$ with $\nu_{D^\rho}=0$. We say that $\nu$ is supported by some subset ${\widetilde{\partial}}\subseteq \partial({{\cal T}})^*$ if $\nu({\widetilde{\partial}})=1$. We denote by $\nu'$ the row vector associated to $\nu$. \[cor2\] Every probability measure $\nu$ on $\partial({{\cal T}})^*$ supported on $\partial({{\cal T}})^{{\cal E}}$ satisfies $\nu'P^*=\eta \, \nu'$ and it is a quasi-stationary distribution, that is it satisfies $$\

Finally, let $ \delta_i\in \partial(\zeta^{{\cal E}})$ for $ i=0,.. ,j$. We have $ $ \begin{aligned } \nonumber & { } & \lim\limits_{n\to \infty}{\mathbb{P}}_{\delta_0 } (Y_i=\delta_i, i=1,.. ,j \, | \, \zeta > n)\\ \nonumber & { } & = \lim\limits_{n\to \infty } { \mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,.. ,j) \frac{{\mathbb{P}}_{\delta_j}(\zeta > n - j) } { { \mathbb{P}}_{\delta_0}(\zeta > n)}\\ \label{rats } & { } & = { \mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,.. ,j) \frac{\varphi_{\delta_j}}{\varphi_{\delta_0}}\eta^{-j}\\ \nonumber & { } & = \prod_{l=0}^{j-1 } \left(\eta^{-1}P_{\delta_l, \delta_{l+1}}\, \frac{\varphi_{\delta_{l+1}}}{\varphi_{\delta_l}}\right).\end{aligned}$$ In (\[rats\ ]) we use $ \lim\limits_{n\to \infty}{\mathbb{P}}(\zeta > n - j)/{\mathbb{P}}(\zeta > n)=\eta^{-j}$, which is a consequence of (\[50e\ ]). Then the resultant role follows. Let $ \nu=(\nu_\delta: \delta\in \partial({{\cal T}})^*)$ be a probability measure on $ \partial({{\cal T}})^*$. If necessary, $ \nu$ will be name with its extension on $ \partial({{\cal T}})$ with $ \nu_{D^\rho}=0$. We say that $ \nu$ is confirm by some subset $ { \widetilde{\partial}}\subseteq \partial({{\cal T}})^*$ if $ \nu({\widetilde{\partial}})=1$. We denote by $ \nu'$ the row vector associate to $ \nu$. \[cor2\ ] Every probability measure $ \nu$ on $ \partial({{\cal T}})^*$ supported on $ \partial({{\cal T}})^{{\cal E}}$ satisfies $ \nu'P^*=\eta \, \nu'$ and it is a quasi - stationary distribution, that is it meet $ $ \

Finwlly, let $\delta_i\in \partixl(\zeta^{{\cal E}})$ for $i=0,..,j$. We iave $$\befin{aligndd} \nonumber &{}&\lim\limits_{n\to \infty}{\larhbb{P}}_{\eelta_0} (Y_i=\delta_i, i=1,..,j \, | \, \zega>n)\\ \nonumbvr &{}& = \lim\limuts_{n\uo \infty} {\mathbb{P}}_{\delta_0}(Y_i=\delbc_i, i=1,..,j) \rvac{{\machub{P}}_{\delta_j}(\zeta>n-j)} {{\kathbb{P}}_{\delda_0}(\zeta>n)}\\ \label{rads} &{}&={\oachbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,..,j) \frac{\varphi_{\delta_t}}{\varphi_{\cepta_0}}\eta^{-j}\\ \nonumbet &{}& =\proc_{j=0}^{j-1} \lsft(\eta^{-1}P_{\delta_l, \delta_{l+1}}\, \frac{\varphi_{\deltz_{l+1}}}{\varphp_{\delta_l}}\right).\end{alogned}$$ In (\[rats\]) we used $\lim\llmitd_{n\to \infty}{\mathbb{P}}(\zfta>n-j)/{\mathbb{K}}(\aetw>b)=\eta^{-j}$, which ks a consequence of (\[50e\]). Jhen the result follows. Let $\nu=(\nu_\ddlta: \belta\in \parjnql({{\cwn T}})^*)$ be a pribabijity measure on $\parthal({{\cal Y}})^*$. If necessary, $\nu$ wiol be identified with its extension on $\kartial({{\cal T}})$ with $\nu_{D^\rho}=0$. We sqy that $\nu$ hs sjppoftes uy aome skbsxt ${\widetilds{\partial}}\subweteq \partial({{\cal T}})^*$ of $\bu({\widetilde{\padtial}})=1$. Re denote by $\nu'$ the row vector associated to $\nu$. \[cor2\] Every probabilitt measure $\nu$ on $\partiwl({{\cal T}})^*$ stpported on $\partial({{\cal T}})^{{\cal E}}$ satisfies $\nu'P^*=\eta \, \ng'$ and kt nw a qjqsl-stationary distribution, that is it satisfies $$\

Finally, let $\delta_i\in \partial(\zeta^{{\cal E}})$ for $i=0,..,j$. $$\begin{aligned} &{}&\lim\limits_{n\to \infty}{\mathbb{P}}_{\delta_0} i=1,..,j \, | \lim\limits_{n\to {\mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,..,j) \frac{{\mathbb{P}}_{\delta_j}(\zeta>n-j)} \label{rats} &{}&={\mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,..,j) \nonumber &{}& =\prod_{l=0}^{j-1} \left(\eta^{-1}P_{\delta_l, \delta_{l+1}}\, \frac{\varphi_{\delta_{l+1}}}{\varphi_{\delta_l}}\right).\end{aligned}$$ (\[rats\]) we used $\lim\limits_{n\to \infty}{\mathbb{P}}(\zeta>n-j)/{\mathbb{P}}(\zeta>n)=\eta^{-j}$, which is a consequence of (\[50e\]). Then the follows. Let $\nu=(\nu_\delta: \delta\in \partial({{\cal T}})^*)$ be a probability measure on $\partial({{\cal T}})^*$. necessary, will identified its extension on $\partial({{\cal T}})$ with $\nu_{D^\rho}=0$. We say that $\nu$ is supported by some subset \partial({{\cal T}})^*$ if $\nu({\widetilde{\partial}})=1$. We denote by $\nu'$ row vector associated to \[cor2\] Every probability measure $\nu$ $\partial({{\cal supported on T}})^{{\cal satisfies \, \nu'$ and is a quasi-stationary distribution, that is it satisfies $$\

Finally, let $\delta_i\in \partial(\Zeta^{{\cal E}})$ foR $i=0,..,j$. We HavE $$\beGiN{aliGned} \Nonumber &{}&\lim\limITs_{n\tO \infty}{\mathbb{P}}_{\delta_0} (Y_i=\deLta_i, i=1,..,J \, | \, \zETa>n)\\ \nONuMber &{}& = \lIm\limitS_{N\tO \INftY} {\mAtHbb{p}}_{\dELtA_0}(Y_i=\deLta_I, i=1,..,j) \frac{{\Mathbb{P}}_{\delTa_j}(\ZeTa>n-j)} {{\mathbb{P}}_{\dELtA_0}(\zeta>n)}\\ \labeL{raTs} &{}&={\mathbb{P}}_{\delTa_0}(Y_I=\delta_I, i=1,..,J) \frAC{\varpHi_{\dElta_j}}{\Varphi_{\DElta_0}}\etA^{-j}\\ \nonumbeR &{}& =\pROd_{l=0}^{j-1} \leFT(\eta^{-1}P_{\deLTA_l, \DeltA_{l+1}}\, \frac{\varphi_{\delta_{L+1}}}{\VaRPhi_{\delta_l}}\right).\End{aliGnED}$$ IN (\[RAts\]) We uSed $\lim\limiTs_{N\to \inFTy}{\mathbB{p}}(\zETA>N-j)/{\mAThbb{P}}(\zeta>n)=\eta^{-J}$, which is a coNSeqUence oF (\[50e\]). theN The resUlt foLlOWs. LEt $\nu=(\nu_\delta: \DeltA\in \partiaL({{\cal T}})^*)$ bE A probabILity meaSure on $\ParTiaL({{\cal t}})^*$. if NeCesSaRY, $\nu$ WIlL be IDenTified wiTh ItS exteNsioN ON $\PArtiAl({{\cAl T}})$ wIth $\nu_{d^\rho}=0$. We say that $\Nu$ iS supPOrtEd by sOme suBset ${\WiDetilDe{\partIal}}\suBsEteq \partial({{\cal T}})^*$ If $\nu({\Widetilde{\ParTiAl}})=1$. WE dEnote BY $\nu'$ the Row VecTor assoCiated tO $\Nu$. \[cOr2\] eVERy Probability measure $\Nu$ ON $\PaRtial({{\cal t}})^*$ suppoRTeD oN $\Partial({{\cAl t}})^{{\caL E}}$ saTISfies $\Nu'P^*=\eTA \, \nU'$ and it is A quasi-STaTiOnary diStRibutiOn, ThaT is It satISfieS $$\

Finally, let $\delta_i\i n \partia l(\ze ta^ {{\ ca l E} })$for $i=0,..,j$ . Wehave $$\begin{aligned} \non um b er & { }& \lim\ limits_ { n\ t o \i nf ty }{\ ma t hb b{P}} _{\ delta_0 } (Y_i=\de lta _i , i=1,..,j \ , | \, \zeta> n)\ \ \nonumber&{} & = \l im \li m its_{ n\t o \in fty} { \ mathbb {P}}_{\de lt a _0}(Y_ i =\delta _ i ,i=1, ..,j) \frac{{\mat h bb { P}}_{\delta_j} (\zeta >n - j) } {{\ mat hbb{P}}_{\ de lta_0 } (\zeta> n )} \ \ \la b el{rats} &{}& ={\mathbb{P } }_{ \delta _0 }(Y _ i=\del ta_i, i = 1,. .,j) \frac{ \var phi_{\del ta_j}} { \varphi _ {\delta _0}}\e ta^ {-j }\\\ no nu mbe r& {}& =\ pro d _{l =0}^{j-1 }\l eft(\ eta^ { - 1 } P_{\ del ta_l , \de lta_{l+1}}\,\fr ac{\ v arp hi_{\ delta _{l+ 1} }}{\v arphi_ {\del ta _l}}\right).\en d{al igned}$$In(\ [ra ts \]) w e used$\l im\ limits_ {n\to \ i nft y} { \ m at hbb{P}}(\zeta>n-j) /{ \ m at hbb{P}}( \zeta> n )= \e t a^{-j}$, w hic h is a cons eque n ce of (\[5 0e\]). Th en the re su lt fol lo ws. L et $\ n u=(\ nu_\de lta: \de lta\i n \partial({{\c a l T}})^*)$ be ap r ob a bili tymeasure on$\pa r tial ({{\ c al T} } )^*$. If n ec e ss a ry, $\nu$ will be i de ntifie d wit h its extensi on on $\pa r t i al({{\ca l T} } )$ with $\nu_{D^\ rho}= 0$. We say that $\n u$ is support ed by som e subset $ {\w ide til de{ \ p ar tial}}\subset e q \pa rt ial({{\ cal T}})^* $ i f $ \nu ({\ wi detilde{\ partial} }) =1 $. W e d enote by $\nu' $the r owvecto r assoc iated to$\ nu $ . \[cor2\ ] E v e ry p ro ba bili tyme asure $\n u $ o n $\par tial({{\c alT }})^ *$ s upporte d on $\partia l( {{\cal T}} )^ {{\ cal E} } $ satisfi es $\nu'P^*=\eta \, \nu ' $ and i t i s a q uasi -stationa rydistri but i on, th at isit sa ti sfi e s $$\

Finally, let_$\delta_i\in _\partial(\zeta^{{\cal E}})$ for $i=0,..,j$._We have_$$\begin{aligned} \nonumber &{}&\lim\limits_{n\to_\infty}{\mathbb{P}}_{\delta_0} (Y_i=\delta_i, i=1,..,j_\,_| \, \zeta>n)\\ \nonumber &{}&_= \lim\limits_{n\to \infty} {\mathbb{P}}_{\delta_0}(Y_i=\delta_i, i=1,..,j) \frac{{\mathbb{P}}_{\delta_j}(\zeta>n-j)} {{\mathbb{P}}_{\delta_0}(\zeta>n)}\\ \label{rats} &{}&={\mathbb{P}}_{\delta_0}(Y_i=\delta_i,_i=1,..,j) \frac{\varphi_{\delta_j}}{\varphi_{\delta_0}}\eta^{-j}\\ \nonumber &{}& =\prod_{l=0}^{j-1} \left(\eta^{-1}P_{\delta_l, \delta_{l+1}}\, \frac{\varphi_{\delta_{l+1}}}{\varphi_{\delta_l}}\right).\end{aligned}$$_In (\[rats\]) we_used_$\lim\limits_{n\to \infty}{\mathbb{P}}(\zeta>n-j)/{\mathbb{P}}(\zeta>n)=\eta^{-j}$, which is a consequence of (\[50e\]). Then the result follows. Let $\nu=(\nu_\delta: \delta\in_\partial({{\cal_T}})^*)$ be_a_probability_measure on $\partial({{\cal T}})^*$. If_necessary, $\nu$ will be identified_with its_extension on $\partial({{\cal T}})$ with $\nu_{D^\rho}=0$. We say_that_$\nu$ is supported_by some subset ${\widetilde{\partial}}\subseteq \partial({{\cal T}})^*$ if $\nu({\widetilde{\partial}})=1$. We_denote by $\nu'$ the row vector_associated to $\nu$. \[cor2\]_Every_probability_measure $\nu$ on $\partial({{\cal_T}})^*$ supported on $\partial({{\cal T}})^{{\cal E}}$_satisfies $\nu'P^*=\eta \, \nu'$ and it_is a quasi-stationary distribution, that is it_satisfies $$\

ni, Y. Shtanov, JCAP **0311**, 014 (2003). --- abstract: 'The zero-temperature Glauber dynamics of the random-field Ising model describes various ubiquitous phenomena such as avalanches, hysteresis, and related critical phenomena. Here, for a model on a random graph with a special initial condition, we derive exactly an evolution equation for an order parameter. Through a bifurcation analysis of the obtained equation, we reveal a new class of cooperative slow dynamics with the determination of critical exponents.' author: - Hiroki Ohta - 'Shin-ichi Sasa' title: 'A universal form of slow dynamics in zero-temperature random-field Ising model' --- [Department of Pure and Applied Sciences, University of Tokyo, 3-8-1 Komaba Meguro-ku, Tokyo 153-8902, Japan]{} Slow dynamical behaviors caused by cooperative phenomena are observed in various many-body systems. In addition to well-studied examples such as critical slowing down [@H-H], phase ordering kinetics [@Bray], and slow relaxation in glassy systems [@Cavagna], seemingly different phenomena from these examples have also been discovered successively. In order to judge whether or not an observed phenomenon is qualitatively new, one needs to determine a universality class including the phenomenon. In this context, it is significant to develop a theoretical method for classifying slow dynamics. Here, let us recall a standard procedure for classifying equilibrium critical phenomena. First, for an order parameter $m$ of a mean-field model, a qualitative change in the solutions of a self-consistent equation $m={\cal F}(m)$ is investigated; then, the differences between the results of the mean-field model and finite-dimensional systems are studied by, for example, a renormalization group method. On the basis of this success, an analysis of the dynamics of a typical mean-field model is expected to be a first step toward determining a universality class of slow dynamics. As an example, in the fully connected Ising model with Glauber dynamics, an evolution equation for the magnetization, $\partial_t m={\cal G}(m)$, can be derived exactly. The analysis of this equation reveals that the critical behavior is described by a pitchfork bifurcation in the dynamical system theory [@Gucken]. As another example, an evolution equation for a time-correlation function and a response function was derived exactly for the fully connected spherical $p

ni, Y. Shtanov, JCAP * * 0311 * *, 014 (2003). --- abstract:' The zero - temperature Glauber dynamics of the random - field Ising model describes diverse omnipresent phenomena such as avalanches, hysteresis, and relate critical phenomenon. Here, for a model on a random graph with a special initial circumstance, we derive exactly an evolution equality for an decree parameter. Through a bifurcation psychoanalysis of the obtained equation, we reveal a new course of cooperative slow moral force with the determination of critical exponent.' writer: - Hiroki Ohta -' Shin - ichi Sasa' title:' A universal form of dull dynamics in zero - temperature random - field Ising model' --- [ Department of Pure and Applied Sciences, University of Tokyo, 3 - 8 - 1 Komaba Meguro - ku, Tokyo 153 - 8902, Japan ] { } dense dynamical behaviors caused by cooperative phenomena are note in various many - body systems. In addition to well - studied example such as critical deceleration down [ @H - H ], phase ordering kinetics [ @Bray ], and slow relaxation in glassy systems [ @Cavagna ], seemingly different phenomenon from these example have also been discovered successively. In decree to pronounce whether or not an observed phenomenon is qualitatively new, one needs to determine a universality class including the phenomenon. In this context, it is significant to modernize a theoretical method for classifying slow dynamics. Here, let us recall a standard routine for classifying equilibrium critical phenomena. First, for an orderliness argument $ m$ of a average - field model, a qualitative change in the solutions of a self - consistent equation $ m={\cal F}(m)$ is investigated; then, the dispute between the results of the mean - field model and finite - dimensional system are studied by, for example, a renormalization group method. On the basis of this success, an psychoanalysis of the dynamics of a typical mean - field mannequin is expected to be a first step toward determining a universality class of dense dynamics. As an exemplar, in the fully connected Ising model with Glauber dynamics, an evolution equality for the magnetization, $ \partial_t m={\cal G}(m)$, can be derived exactly. The psychoanalysis of this equation reveals that the critical behavior is identify by a pitchfork bifurcation in the dynamical system theory [ @Gucken ]. As another exercise, an evolution equation for a time - correlation function and a response affair was derive precisely for the fully connected spherical $ p

ni, J. Shtanov, JCAP **0311**, 014 (2003). --- abstragt: 'The zero-tempetarure Gnauber dynamicr of the random-field Ising mldwl dewcribes various ubiquigous phenlmena suxh aw avalanches, hystercfis, zkd reneted critical pmenomena. Hese, for a model ov c random graph with a special initiaj condiyiln, we derive evactkr an vvilution equation for an order parameuer. Through a bifutcation analysis of the obhainfd equation, we revfal a new coass if cooperatixe slow dynamics with jhe determination of critical exoonencs.' author: - Hitoii Mhta - 'Shin-icii Sasw' title: 'A uniyvrsal fmrm of xlow dynamics ln zeco-tenperature random-field Ising model' --- [Departient of Pgrz and Applied Sciencew, Ynivetsity of Gikyu, 3-8-1 Iokaga Megkro-iu, Tokyo 153-8902, Napan]{} Slow dtnamical behaviors vatwed by cooperztive [hqnomena are observed in various many-bodj syatems. In addition to weol-studied examples sufh as crieical slowing down [@H-H], phase ordering kinetics [@Braf], and rloc relabqtlon in glassy systems [@Cavagna], seemingly diffewsnu pmenomena from thcse examples have sldo feen discovergd succzasjvely. In order to uudge wrethee or not wn onserved phenomenon is qualiratively new, ine needs to deterline a univzrsalijy claxs including the phenomznon. Ih this contfxt, it is rignificant to ddveko[ a theoretical method for classifymng skow dynxmicx. Here, jet us recwll a standard procedure for elasshfying equllibrium critical phenomena. First, for an ordet pdraketer $m$ jf a kean-field modql, a qualitatire changz in tfe solutiohs of a self-consiseent equation $l={\cal F}(m)$ is iivestigatqd; tyen, rhe difwdrences betweem the resllus of the nean-field model anv fivjte-dimensional wysuens are studied by, fow vxakpja, a renormalhzatkon broup method. On bhe basos of this success, at anzlysis of the dynakigs of a tipical mewn-field model is expected to be a ficst stxp towsrd determining a universality clzss of sllw qynamics. As ag exqmple, in the fully connected Ising model with Glaubxr dynamics, an evolutiob equation for the ksgnetizatioi, $\partyal_t m={\cal G}(m)$, can be derived ezactly. The analysls of this equation rebeals dhat hhe critical behavior is described by a pitchfork bifurcation in the dynanical system theorg [@Guvken]. Ds cnother exal'le, an evolution vquation for a time-correlation fnnction ang c response function was derivrd exactly for jhe fully connected spheridal $p

ni, Y. Shtanov, JCAP **0311**, 014 (2003). 'The Glauber dynamics the random-field Ising such avalanches, hysteresis, and critical phenomena. Here, a model on a random graph a special initial condition, we derive exactly an evolution equation for an order Through a bifurcation analysis of the obtained equation, we reveal a new class cooperative dynamics the of critical exponents.' author: - Hiroki Ohta - 'Shin-ichi Sasa' title: 'A universal form of slow in zero-temperature random-field Ising model' --- [Department of and Applied Sciences, University Tokyo, 3-8-1 Komaba Meguro-ku, Tokyo Japan]{} dynamical behaviors by phenomena observed in various systems. In addition to well-studied examples such as critical slowing down [@H-H], phase ordering kinetics [@Bray], and relaxation in [@Cavagna], seemingly phenomena these have also been In order to judge whether or phenomenon is qualitatively new, one needs to determine universality class the phenomenon. In this context, it significant to develop a theoretical method for classifying dynamics. Here, let us recall a standard procedure for classifying equilibrium critical phenomena. First, for parameter $m$ of a model, a qualitative in solutions a equation $m={\cal is investigated; then, the differences between the results of the mean-field and finite-dimensional systems are studied by, for example, a renormalization On basis of this an analysis of the of typical mean-field model is be first a class slow dynamics. As an in the fully connected Ising with Glauber dynamics, an $\partial_t m={\cal G}(m)$, can be derived exactly. The of this equation reveals that the critical is described by a pitchfork bifurcation in the dynamical system theory [@Gucken]. another example, equation for a time-correlation function and a response was derived exactly for fully connected spherical $p

ni, Y. Shtanov, JCAP **0311**, 014 (2003). --- abstract: 'The Zero-temperAture glaUbeR dYnamIcs oF the random-fielD isinG model describes various UbiquItOUs phENoMena sUch as avALaNCHes, HyStEreSiS, AnD relaTed CriticaL phenomena. herE, fOr a model on a rANdOm graph witH a sPecial initiaL coNditioN, wE deRIve exActLy an eVolutiON equatIon for an oRdER paramETer. ThroUGH a BifuRcation analysis of THe OBtained equatioN, we revEaL A nEW ClaSs oF cooperatiVe Slow dYNamics wITh THE DetERmination of crItical exponENts.' Author: - hiRokI ohta - 'ShIn-ichI SASa' tItle: 'A univerSal fOrm of slow DynamiCS in zero-TEmperatUre ranDom-FieLd IsINg MoDel' --- [dePArtMEnT of pUre And AppliEd scIenceS, UniVERSIty oF ToKyo, 3-8-1 KOmaba meguro-ku, Tokyo 153-8902, japAn]{} SlOW dyNamicAl behAvioRs CauseD by cooPeratIvE phenomena are obServEd in varioUs mAnY-boDy SysteMS. In addItiOn tO well-stUdied exAMplEs SUCH aS critical slowing doWn [@h-h], PhAse orderIng kinETiCs [@bRay], and slOw RelAxatION in glAssy SYsTems [@CavaGna], seeMInGlY differEnT phenoMeNa fRom These EXampLes havE also beeN discOVered successivELy. In order to juDGe WHEtHEr or Not An observed pHenoMEnon Is quALiTatIVely nEw, one NeEDs TO determine a universaLiTy clasS inclUding the phenoMenon. In thiS CONtext, it iS sigNIfICant to develop a TheorEtical methOD for clasSifyiNg slow dyNamics. HerE, LEt us recaLl a StaNdaRd pROCeDure for classiFYIng eQuIlibriuM crItical pHenOmeNa. FIrsT, fOr an order ParameteR $m$ Of A mEaN-fiEld moDEl, a qualiTaTivE cHanGe in tHE solutIons oF a seLf-CoNSisTent equATiON $M={\cal f}(m)$ Is InveStiGaTed; thEn, thE DifFerenceS between tHe rESultS oF tHe mean-fIeld model and fInIte-dimensiOnAl sYstems ARE studied By, for example, a renormalizATion groUp mEthod. on thE basis of tHis SuccesS, an ANalysiS of the DynamIcS of A TYpicaL MEaN-fiElD model is exPECteD to be A fIrst Step towArd determining a uniVErsAlity class of sLow DynaMICs. as aN ExAMplE, iN The FULly connected IsiNg model witH GLAuBer dynamicS, An eVoLution eQuation For thE MagnetiZation, $\parTial_t m={\cal g}(m)$, Can bE DEriVed exactly. the analySis of this EQuatiON rEvealS thAt the cRiTicAl behAvior iS DesCribeD by a piTcHfork bIfurcAtIon in the Dynamical system theory [@GuCken]. As AnothEr eXample, an eVolUTioN equation For a Time-correlAtiOn fUnctiOn aND a resPonsE FuNctIOn was DeriVEd exactly FOr The FULlY connected sPHERicAl $p

ni, Y. Shtanov, JCAP **031 1**, 014 ( 2003) . --- a bstr act: 'The zero-tem p erat ure Glauber dynamics o f the r a ndom - fi eld I sing mo d el d esc ri be s v ar i ou s ubi qui tous ph enomena su chas avalanches, hy steresis,and related cri tic al phe no men a . Her e,for a model on a r andom gra ph with a special i ni tial condition, we de r iv e exactly an ev olutio ne qu a t ion fo r an order p arame t er. Thr o ug h a bi f urcation anal ysis of the obt ainedeq uat i on, we reve al a n ew class of coo perativeslow d y namicsw ith the deter min ati on o f c ri tic al exp o ne nts . ' a uthor: - H ir oki O hta- ' S hin- ich i Sa sa' t itle: 'A univ ers al f o rmof sl ow dy nami cs in z ero-te mpera tu re random-field Isi ng model' -- - [D ep artme n t of P ure an d Appli ed Scie n ces ,U n i ve rsity of Tokyo, 3- 8- 1 Ko maba Meg uro-ku , T ok y o 153-89 02 , J apan ] { } Sl ow d y na mical be havior s c au sed byco operat iv e p hen omena areobserv ed in va rious many-body syst e ms. In additi o nt o w e ll-s tud ied example s su c h as cri t ic als lowin g dow n[ @H - H], phase orderingki netics [@Br ay], and slow relaxatio n i n glassy sys t em s [@Cavagna], s eemin gly differ e nt pheno menafrom the se exampl e s have al sobee n d isc o v er ed successive l y . In o rder to ju dge whe the r o r n otan observed phenome no nis q ual itati v ely new, o nene eds to d e termin e a u nive rs al i tyclass i n cl u d ingth ephen ome no n. In thi s co ntext,it is sig nif i cant t odevelop a theoretica lmethod for c las sifyin g slow dyn amics. Here, let us re c all a s tan dardproc edure for cl assify ing equili briumcriti ca l p h e nomen a . F irs t, for an or d e r p arame te r $m $ of amean-field model,a qu alitative cha nge int h esol u ti o nsof a s e l f-consistent eq uation $m= {\ c al F}(m)$ is inv es tigated ; then, thed ifferen ces betwe en the re su ltso f th e mean-fie ld model and fini t e-dim e ns ional sy stemsar e s tudie d by,f orexamp le, are normal izati on group m ethod. On the basis ofthis s ucces s,an analys iso f t he dynami cs o f a typica l m ean -fiel d m o del i s ex p ec ted to be a f i rst stept ow ard d et ermining au n i ver salit y c l ass of slo w dynamics. As a n example, in t he f u l lycon n ecte dIsing model wi thGl a u ber dyna mi cs, an evol ution eq ua t ion f or the magne tizatio n , $ \ partia l_tm={ \cal G}(m )$, c a n be de ri ve d exact ly.Th e anal ysis o f thi s equation reveals that t he cr i tic al be ha vior is desc ribed by a pitchforkbifurc atio n inthe dyn am ical s yst em theory [@ G ucken]. A s ano ther ex am ple, an evolu tion e quati on f or atime-corr e l at i on f u nct ionand a r espo nse funct i on was d eri v ed exac tl y f o r the f u ll y connectedsph erica l $p

ni, Y._Shtanov, JCAP_**0311**, 014 (2003). --- abstract:_'The zero-temperature_Glauber_dynamics of_the_random-field Ising model_describes various ubiquitous_phenomena such as avalanches,_hysteresis, and related_critical_phenomena. Here, for a model on a random graph with a special initial condition,_we_derive exactly_an_evolution_equation for an order parameter._Through a bifurcation analysis of_the obtained_equation, we reveal a new class of cooperative_slow_dynamics with the_determination of critical exponents.' author: - Hiroki Ohta - 'Shin-ichi Sasa' title: 'A_universal form of slow dynamics in_zero-temperature random-field Ising_model' --- [Department_of_Pure and Applied Sciences,_University of Tokyo, 3-8-1 Komaba Meguro-ku,_Tokyo 153-8902, Japan]{} Slow dynamical behaviors caused_by cooperative phenomena are observed in various_many-body systems. In addition to well-studied_examples such as critical slowing_down [@H-H],_phase ordering kinetics [@Bray], and_slow relaxation in_glassy systems_[@Cavagna], seemingly different_phenomena from these examples have also_been discovered successively._In order to judge whether or_not_an observed phenomenon_is_qualitatively_new, one_needs to determine_a_universality class_including_the phenomenon. In this context, it_is_significant to develop a theoretical method for_classifying slow dynamics. Here, let_us_recall a standard procedure_for classifying equilibrium critical phenomena._First, for an order parameter $m$_of a_mean-field model,_a qualitative change in the solutions of a self-consistent equation $m={\cal_F}(m)$ is investigated; then, the differences_between the results of_the mean-field_model_and finite-dimensional systems_are_studied by,_for example, a renormalization group method. On_the basis_of this success, an analysis of_the dynamics of a_typical_mean-field model is expected to be_a first step toward determining a_universality class of slow dynamics. As_an_example,_in the fully connected Ising_model with Glauber dynamics, an evolution_equation for the_magnetization, $\partial_t m={\cal G}(m)$, can be derived_exactly._The analysis of this equation reveals_that_the critical behavior is described by_a_pitchfork_bifurcation in the dynamical system_theory [@Gucken]. As another example, an_evolution equation for a time-correlation function and a response_function was derived_exactly for the fully connected_spherical_$p

1(E\cap B(x,r)\setminus X(x,L,\alpha))}{2r}=0\quad\text{for all }\alpha\in(0,\pi/2],$$ where $X(x,L,\alpha)$ denotes the cone of points $y\in{\mathbb{R}}^2$ such that the line $\{x+t(y-x):t\in{\mathbb{R}}\}$ meets the line $L$ at an angle at most $\alpha$ (see Figure 1.2). For a contemporary, self-contained presentation of these results, see [@Falconer]. {width=".7\textwidth"} Let $(\mathbb{X},\mathcal{M})=({\mathbb{R}}^2,\mathcal{B}_{{\mathbb{R}}^2})$, let $\mathcal{N}$ be the collection of rectifiable curves in ${\mathbb{R}}^2$, and let $\mathscr{F}$ denote the collection of Radon measures (locally finite, Borel regular measures) $\mu$ on ${\mathbb{R}}^2$ **such that** $\mu\ll{\mathcal{H}}^1$. Morse and Randolph [@MR] proved that if $\mu\in\mathscr{F}$, then $$\mu^1_{{rect}}\equiv\mu_\mathcal{N}=\mu{\hbox{ {\vrule height.22cm}{\leaders\hrule\hskip.2cm} }}\left\{x\in{\mathbb{R}}^2: \lim_{r\downarrow 0}\frac{\mu(B(x,r))}{2r}\text{ exists, and }0<\lim_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}<\infty\right\},$$ $$\mu^1_{{pu}}\equiv\mu_\mathcal{N}^\perp=\mu{\hbox{ {\vrule height.22cm}{\leaders\hrule\hskip.2cm} }}\left\{x\in{\mathbb{R}}^2: \liminf_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}\leq \frac{100}{101}\limsup_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}\right\}.$$ Moore [@Moore] extended Morse and Randolph’s density characterizations of $\mu^

1(E\cap B(x, r)\setminus X(x, L,\alpha))}{2r}=0\quad\text{for all } \alpha\in(0,\pi/2],$$ where $ X(x, L,\alpha)$ denotes the cone of points $ y\in{\mathbb{R}}^2 $ such that the line $ \{x+t(y - x):t\in{\mathbb{R}}\}$ meets the course $ L$ at an slant at most $ \alpha$ (see name 1.2). For a contemporaneous, self - contained presentation of these results, watch [ @Falconer ]. ! [ the cone $ X(x, L,\alpha)$](cone){width=".7\textwidth " } Let $ (\mathbb{X},\mathcal{M})=({\mathbb{R}}^2,\mathcal{B}_{{\mathbb{R}}^2})$, let $ \mathcal{N}$ be the solicitation of reparable curves in $ { \mathbb{R}}^2 $, and let $ \mathscr{F}$ announce the collection of Radon measures (locally finite, Borel unconstipated measure) $ \mu$ on $ { \mathbb{R}}^2 $ * * such that * * $ \mu\ll{\mathcal{H}}^1$. Morse and Randolph [ @MR ] proved that if $ \mu\in\mathscr{F}$, then $ $ \mu^1_{{rect}}\equiv\mu_\mathcal{N}=\mu{\hbox { { \vrule height.22cm}{\leaders\hrule\hskip.2 cm } } } \left\{x\in{\mathbb{R}}^2: \lim_{r\downarrow 0}\frac{\mu(B(x, r))}{2r}\text { exists, and } 0<\lim_{r\downarrow 0 } \frac{\mu(B(x, r))}{2r}<\infty\right\},$$ $ $ \mu^1_{{pu}}\equiv\mu_\mathcal{N}^\perp=\mu{\hbox { { \vrule height.22cm}{\leaders\hrule\hskip.2 cm } } } \left\{x\in{\mathbb{R}}^2: \liminf_{r\downarrow 0 } \frac{\mu(B(x, r))}{2r}\leq \frac{100}{101}\limsup_{r\downarrow 0 } \frac{\mu(B(x, r))}{2r}\right\}.$$ Moore [ @Moore ] extended Morse and Randolph ’s density characterizations of $ \mu^

1(E\cao B(x,r)\setminus X(x,L,\alpha))}{2r}=0\qmad\text{for all }\alpha\in(0,\pm/2],$$ where $X(x,L,\alphx)$ denotes the cone of points $y\un{\matybb{R}}^2$ such that the lind $\{x+t(y-x):t\in{\lathbb{R}}\}$ neetw the line $L$ at an angle zb mosc $\elpha$ (see Figurg 1.2). For a condemporary, self-wovtcined presentation of these results, fee [@Falvojer]. {width=".7\textwidth"} Let $(\mathbb{X},\mathcal{M})=({\jathbb{R}}^2,\kathcal{B}_{{\mathbn{R}}^2})$, let $\mathcal{N}$ be the colpectlon of rectifiable curves in ${\nathfv{R}}^2$, and let $\mxthscr{F}$ denote the colmection of Radon measures (localuy fiuite, Borel tztulwt measures) $\mn$ on ${\mwthbb{R}}^2$ **such bnat** $\mu\nl{\mathcsl{H}}^1$. Morse and Vandonph [@MR] proved that if $\mu\mn\mathscr{F}$, then $$\mu^1_{{rest}}\equiv\mu_\kachcal{N}=\mu{\hbox{ {\vrule heugyt.22cm}{\lgaderv\hruuw\hsyip.2dm} }}\lsft\{x\in{\latibb{R}}^2: \lim_{r\dosnarrow 0}\frax{\mu(B(x,r))}{2r}\text{ exists, snq }0<\lim_{r\downarros 0} \fras{\mt(B(x,r))}{2r}<\infty\right\},$$ $$\mu^1_{{pu}}\equiv\mu_\mathcal{N}^\perp=\ku{\hgox{ {\vrule height.22cm}{\leadees\hrule\hskip.2cm} }}\left\{x\ij{\mathbb{R}}^2: \jiminf_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}\leq \frac{100}{101}\limsup_{r\downarsow 0} \hrxc{\mb(N(w,r))}{2r}\rkthh\}.$$ Moore [@Moore] extended Morse and Randolph’s degaiuy bharacterizations of $\mu^

1(E\cap B(x,r)\setminus X(x,L,\alpha))}{2r}=0\quad\text{for all }\alpha\in(0,\pi/2],$$ where $X(x,L,\alpha)$ cone points $y\in{\mathbb{R}}^2$ that the line at angle at most (see Figure 1.2). a contemporary, self-contained presentation of these see [@Falconer]. {width=".7\textwidth"} Let $(\mathbb{X},\mathcal{M})=({\mathbb{R}}^2,\mathcal{B}_{{\mathbb{R}}^2})$, let $\mathcal{N}$ be the collection of curves in ${\mathbb{R}}^2$, and let $\mathscr{F}$ denote the collection of Radon measures (locally Borel measures) on **such that** $\mu\ll{\mathcal{H}}^1$. Morse and Randolph [@MR] proved that if $\mu\in\mathscr{F}$, then $$\mu^1_{{rect}}\equiv\mu_\mathcal{N}=\mu{\hbox{ {\vrule height.22cm}{\leaders\hrule\hskip.2cm} }}\left\{x\in{\mathbb{R}}^2: 0}\frac{\mu(B(x,r))}{2r}\text{ exists, and }0<\lim_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}<\infty\right\},$$ $$\mu^1_{{pu}}\equiv\mu_\mathcal{N}^\perp=\mu{\hbox{ {\vrule }}\left\{x\in{\mathbb{R}}^2: \liminf_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}\leq 0} \frac{\mu(B(x,r))}{2r}\right\}.$$ Moore [@Moore] extended and density characterizations $\mu^

1(E\cap B(x,r)\setminus X(x,L,\alpha))}{2r}=0\qUad\text{for All }\alPha\In(0,\pI/2],$$ wHere $x(x,L,\aLpha)$ denotes the COne oF points $y\in{\mathbb{R}}^2$ such tHat thE lINe $\{x+t(Y-X):t\In{\matHbb{R}}\}$ meeTS tHE LinE $L$ At An aNgLE aT most $\AlpHa$ (see FiGure 1.2). For a coNteMpOrary, self-conTAiNed presentAtiOn of these resUltS, see [@FaLcOneR]. {WIdth=".7\teXtwidth"} LeT $(\mAThbb{X},\mAThcal{M})=({\mATHbB{R}}^2,\maThcal{B}_{{\mathbb{R}}^2})$, let $\mAThCAl{N}$ be the collecTion of ReCTiFIAblE cuRves in ${\mathBb{r}}^2$, and lET $\mathscR{f}$ dENOTe tHE collection of radon measurES (loCally fInIte, bOrel reGular MeASurEs) $\mu$ on ${\mathbB{R}}^2$ **suCh that** $\mu\lL{\mathcAL{H}}^1$. Morse ANd RandoLph [@MR] pRovEd tHat iF $\Mu\In\MatHsCR{F}$, tHEn $$\Mu^1_{{rECt}}\eQuiv\mu_\maThCaL{N}=\mu{\hBox{ {\vRULE HeigHt.22cM}{\leaDers\hRule\hskip.2cm} }}\leFt\{x\In{\maTHbb{r}}^2: \lim_{r\DownaRrow 0}\FrAc{\mu(B(X,r))}{2r}\texT{ exisTs, And }0<\lim_{r\downarroW 0} \fraC{\mu(B(x,r))}{2r}<\inFty\RiGht\},$$ $$\Mu^1_{{Pu}}\equIV\mu_\matHcaL{N}^\pErp=\mu{\hbOx{ {\vrule HEigHt.22CM}{\LEaDers\hrule\hskip.2cm} }}\leFt\{X\IN{\mAthbb{R}}^2: \liMinf_{r\dOWnArROw 0} \frac{\mu(b(x,R))}{2r}\lEq \frAC{100}{101}\LimsuP_{r\doWNaRrow 0} \frac{\Mu(B(x,r))}{2r}\RIgHt\}.$$ moore [@MoOrE] extenDeD MoRse And RaNDolpH’s densIty charaCteriZAtions of $\mu^

1(E\cap B(x,r)\setminus X( x,L,\alpha ))}{2 r}= 0\q ua d\te xt{f or all }\alpha \ in(0 ,\pi/2],$$ where $X(x, L,\al ph a )$ d e no tes t he cone of p oin ts $ y\i n{ \ ma thbb{ R}} ^2$ suc h that the li ne $\{x+t(y-x) : t\ in{\mathbb {R} }\}$ meets t heline $ L$ at an an gle at m ost $\ a lpha$(see Figu re 1.2).F or a co n t em pora ry, self-containe d p r esentation ofthesere s ul t s , s ee[@Falconer ]. { width=".7\t e xtw idth"} Let $(\mat hbb{X }, \ mat hcal{M})=({ \mat hbb{R}}^2 ,\math c al{B}_{ { \mathbb {R}}^2 })$ , l et $ \ ma th cal {N } $ b e t hec oll ection o fre ctifi able c u r vesin${\m athbb {R}}^2$, andlet $\m a ths cr{F} $ den oteth e col lectio n ofRa don measures (l ocal ly finite , B or elre gular measur es) $\ mu$ on${\math b b{R }} ^ 2 $ * *such that** $\mu\ ll { \ ma thcal{H} }^1$.M or se and Rand ol ph[@MR ] prove d th a tif $\mu\ in\mat h sc r{ F}$, th en $$\mu ^1 _{{ rec t}}\e q uiv\ mu_\ma thcal{N} =\mu{ \ hbox{ {\vruleh eight.22cm}{\ l ea d e rs \ hrul e\h skip.2cm} } }\le f t\{x \in{ \ ma thb b {R}}^ 2: \l im _ {r \ downarrow 0}\frac{\ mu (B(x,r ))}{2 r}\text{ exis ts, and }0 < \ l im_{r\do wnar r ow 0} \frac{\mu(B (x,r) )}{2r}<\in f ty\right \},$$ $$\mu^1 _{{pu}}\e q u iv\mu_\m ath cal {N} ^\p e r p= \mu{\hbox{ {\ v r ulehe ight.22 cm} {\leade rs\ hru le\ hsk ip .2cm} }}\ left\{x\ in {\ ma th bb{ R}}^2 : \liminf _{ r\d ow nar row 0 } \frac {\mu( B(x, r) )} { 2r} \leq \f r ac { 1 00}{ 10 1} \lim sup _{ r\dow narr o w 0 } \frac {\mu(B(x, r)) } {2r} \r ig ht\}.$$ Moore [@Moor e] extendedMo rse and R a n dolph’sdensity characterizatio n s of $\ mu^

1(E\cap B(x,r)\setminus_X(x,L,\alpha))}{2r}=0\quad\text{for all_}\alpha\in(0,\pi/2],$$ where $X(x,L,\alpha)$ denotes_the cone_of_points $y\in{\mathbb{R}}^2$_such_that the line_$\{x+t(y-x):t\in{\mathbb{R}}\}$ meets the_line $L$ at an_angle at most_$\alpha$_(see Figure 1.2). For a contemporary, self-contained presentation of these results, see [@Falconer]. {width=".7\textwidth"} Let_$(\mathbb{X},\mathcal{M})=({\mathbb{R}}^2,\mathcal{B}_{{\mathbb{R}}^2})$, let_$\mathcal{N}$_be_the collection of rectifiable curves_in ${\mathbb{R}}^2$, and let $\mathscr{F}$_denote the_collection of Radon measures (locally finite, Borel regular_measures)_$\mu$ on ${\mathbb{R}}^2$_**such that** $\mu\ll{\mathcal{H}}^1$. Morse and Randolph [@MR] proved that_if $\mu\in\mathscr{F}$, then $$\mu^1_{{rect}}\equiv\mu_\mathcal{N}=\mu{\hbox{ {\vrule height.22cm}{\leaders\hrule\hskip.2cm}_}}\left\{x\in{\mathbb{R}}^2: \lim_{r\downarrow 0}\frac{\mu(B(x,r))}{2r}\text{_exists,_and_}0<\lim_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}<\infty\right\},$$ $$\mu^1_{{pu}}\equiv\mu_\mathcal{N}^\perp=\mu{\hbox{_{\vrule height.22cm}{\leaders\hrule\hskip.2cm} }}\left\{x\in{\mathbb{R}}^2: \liminf_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}\leq_\frac{100}{101}\limsup_{r\downarrow 0} \frac{\mu(B(x,r))}{2r}\right\}.$$ Moore [@Moore] extended_Morse and Randolph’s density characterizations of $\mu^

imality assumptions have been introduced. These conditions are weaker than the CFC but do not guarantee model identifiability [@Zhang2013]. On the other hand, the recently introduced sparsest Markov representation (SMR) and frugality assumptions [@forster2015frugal; @Raskutti2013; @van2013ell] provide an alternative that is milder than the CFC and is sufficient to ensure identifiability. The main downside of the [SMR]{} and frugality assumptions relative to the CFC is that the [SMR]{} and frugality assumptions are sufficient conditions for model identifiability only when exhaustive searches over the DAG space are possible [@Raskutti2013], while the CFC is sufficient for polynomial-time algorithms [@Glymour1987; @Spirtes1991; @Spirtes2000] for learning equivalence class of sparse graphs. While the DAG framework is useful in many applications, it is limited since feedback loops are known to often exist (see e.g., [@Richardson1996; @Richardson1995]). Hence, directed graphs with directed cycles [@Spirtes2000] are more appropriate to model such feedback. However learning directed cyclic graphical (DCG) models from data is considerably more challenging than learning DAG models [@Richardson1996; @Richardson1995] since the presence of cycles poses a number of additional challenges and introduces additional non-identifiability. Consequently there has been considerably less work focusing on directed graphs with feedback both in terms of identifiability assumptions and search algorithms. [@Spirtes1995] discussed the CMC, and [@Richardson1996; @Richardson1995] discussed the CFC for DCG models and introduced the polynomial-time cyclic causal discovery (CCD) algorithm [@Richardson1996] for recovering the Markov equivalence class for DCGs. Recently, [@claassen2013learning] introduced the FCI$+$ algorithm for recovering the Markov equivalence class for sparse DCGs, which also assumes the CFC. As with DAG models, the CFC for cyclic models is extremely restrictive since it is more restrictive than the CFC for DAG models. In terms of learning algorithms that do not require the CFC, additional assumptions are typically required. For example [@mooij2011causal] proved identifiability for bivariate Gaussian cyclic graphical models with additive noise which does not require the CFC while many approaches have been studied for learning

imality assumptions have been introduced. These conditions are weaker than the CFC but do not undertake exemplar identifiability   [ @Zhang2013 ]. On the other hand, the recently insert sparse Markov representation (SMR) and frugality assumptions   [ @forster2015frugal; @Raskutti2013; @van2013ell ] provide an option that is meek than the CFC and is sufficient to ensure identifiability. The main downside of the [ SMR ] { }   and frugality assumptions relative to the CFC is that the [ SMR ] { }   and frugality assumption are sufficient conditions for model identifiability entirely when exhaustive search over the DAG space are potential   [ @Raskutti2013 ], while the CFC is sufficient for polynomial - time algorithm   [ @Glymour1987; @Spirtes1991; @Spirtes2000 ] for learn equivalence class of sparse graphs. While the DAG model is useful in many applications, it is limited since feedback loops are known to frequently exist (see e.g.,   [ @Richardson1996; @Richardson1995 ]). Hence, directed graphs with target cycles   [ @Spirtes2000 ] are more appropriate to model such feedback. However learning directed cyclic graphical (DCG) models from datum is well more ambitious than learning DAG models   [ @Richardson1996; @Richardson1995 ] since the presence of cycles poses a number of additional challenge and introduce additional non - identifiability. Consequently there has been considerably less work concenter on directed graphs with feedback both in terms of identifiability assumption and search algorithms. [ @Spirtes1995 ] discussed the CMC, and   [ @Richardson1996; @Richardson1995 ] discussed the CFC for DCG models and introduce the polynomial - time cyclic causal discovery (CCD) algorithm   [ @Richardson1996 ] for recovering the Markov equivalence class for DCGs. Recently,   [ @claassen2013learning ] introduced the FCI$+$ algorithm for recovering the Markov comparison class for sparse DCGs, which also assumes the CFC. As with DAG models, the CFC for cyclic model is highly restrictive since it is more restrictive than the CFC for DAG models. In terms of learning algorithms that do not require the CFC, additional assumptions are typically required. For case   [ @mooij2011causal ] proved identifiability for bivariate Gaussian cyclic graphical models with additive randomness which does not require the CFC while many approaches have been studied for learning

imapity assumptions have betn introduced. These condmtions zre weakdr than the CFC but do not gnarabtee nodel identifiability [@Zfang2013]. On tje other hanv, the recently iifroduced sparacst Mcrjov representajion (SMR) and frugality assgmotnons [@forster2015frugal; @Raskutti2013; @van2013ell] prjvide am wlternative thwt ix milsvr than the CFC and is sufficieht to eisure identifianility. The main downside ov thf [SMR]{} and frugality assumptionw rejqtive to the CFC is that the [SMR]{} ans frugality assumptions are sufwicieut conditiobs fog model idenvifiabplity only whcm exhagstive xearches over bhe DEG space are possible [@Raskntti2013], while the CFC if sufficianc for polynomial-time qltoritvms [@Gnymoje1987; @Soirues1991; @Slirtes2000] foc learning squivalence class of sparse grsprw. While the DAF framqwjrk is useful in many applications, it iv ljmited since feedback liops are known to oftgn exist (sqe e.g., [@Richardson1996; @Richardson1995]). Hence, directed graphs fith vifecuee cycuws [@Dpirtes2000] are more appropriate to model such feqsbscl. However learking directed cycloc gtwphical (DCG) mudels ydoj data is considerwbly mote chaolenging uhan kearning DAG models [@Richardsin1996; @Richardsou1995] sunce the presence lf cycles puses a nukber of additional chaluengss and intrlduces adsktional non-identkfisbhlity. Consequently there hws been cinsiberably uess work socusing oj dircwted graphs with ffedbaek bodh in termd of identifiability assumptions and search akgmrinhms. [@Spirces1995] dixcussed the CIC, and [@Richardspn1996; @Richcrdson1995] discussed the CFR for DCG moqels and intrmfuced the ponynomial-eime cycoic caurxl discovery (CVD) algorinhi [@Ricyardson1996] for recoveving jhs Markov equivaoenxe class for DCBs. Fecqnnly, [@rlaasfan2013learning] ittroaucdc the FCI$+$ algoribhm for recovering the Markmv esuivalence class fpr sparse ECGs, whish also assumrs the CFC. As with DAG kodxls, thr CSC for cyclic models is extremsly restrlctlve since it ys mire restrictnve than the CFC for DAG models. In terms of learning algotithms that do not reamire the CFC, addieional asvumptions are typicaoly required. For txample [@mooij2011causal] provsd idettifiwbility for bivariate Gaussian cyclic graphical models with additive noisw whici qoes not resuirr the CYC chile mwny ep'roaches have beem studied for learning

imality assumptions have been introduced. These conditions than CFC but not guarantee model hand, recently introduced sparsest representation (SMR) and assumptions [@forster2015frugal; @Raskutti2013; @van2013ell] provide an that is milder than the CFC and is sufficient to ensure identifiability. The downside of the [SMR]{} and frugality assumptions relative to the CFC is that [SMR]{} frugality are conditions for model identifiability only when exhaustive searches over the DAG space are possible [@Raskutti2013], while CFC is sufficient for polynomial-time algorithms [@Glymour1987; @Spirtes1991; for learning equivalence class sparse graphs. While the DAG is in many it limited feedback loops are to often exist (see e.g., [@Richardson1996; @Richardson1995]). Hence, directed graphs with directed cycles [@Spirtes2000] are more appropriate model such learning directed graphical models data is considerably than learning DAG models [@Richardson1996; @Richardson1995] of cycles poses a number of additional challenges introduces additional Consequently there has been considerably less focusing on directed graphs with feedback both in of identifiability assumptions and search algorithms. [@Spirtes1995] discussed the CMC, and [@Richardson1996; @Richardson1995] discussed the DCG models and introduced polynomial-time cyclic causal (CCD) [@Richardson1996] recovering Markov equivalence for DCGs. Recently, [@claassen2013learning] introduced the FCI$+$ algorithm for recovering the equivalence class for sparse DCGs, which also assumes the CFC. DAG the CFC for models is extremely restrictive it more restrictive than the DAG In algorithms do require the CFC, additional are typically required. For example proved identifiability for bivariate additive noise which does not require the CFC many approaches have been studied for learning

imality assumptions have beeN introduceD. ThesE coNdiTiOns aRe weAker than the CFC BUt do Not guarantee model identIfiabIlITy [@ZhANg2013]. on the Other haND, tHE RecEnTlY inTrODuCed spArsEst MarkOv represenTatIoN (SMR) and frugaLItY assumptioNs [@fOrster2015frugal; @rasKutti2013; @vAn2013Ell] PRovidE an AlterNative THat is mIlder than ThE cFC and IS sufficIENt To enSure identifiabiliTY. THE main downside oF the [SMr]{} aND fRUGalIty AssumptionS rElatiVE to the Cfc iS THAt tHE [SMR]{} and frugalIty assumptiONs aRe suffIcIenT ConditIons fOr MOdeL identifiabIlitY only when ExhausTIve searCHes over The DAG SpaCe aRe poSSiBlE [@RaSkUTti2013], WHiLe tHE CFc is suffiCiEnT for pOlynOMIAL-timE alGoriThms [@GLymour1987; @Spirtes1991; @spiRtes2000] FOr lEarniNg equIvalEnCe claSs of spArse gRaPhs. While the DAG fRameWork is useFul In ManY aPplicATions, iT is LimIted sinCe feedbACk lOoPS ARe Known to often exist (sEe E.G., [@riChardson1996; @richarDSoN1995]). HENce, direcTeD grAphs WITh dirEcteD CyCles [@SpirTes2000] are MOrE aPpropriAtE to modEl SucH feEdbacK. howeVer leaRning dirEcted CYclic graphical (dcG) models from dATa IS CoNSideRabLy more challEngiNG thaN leaRNiNg Dag modeLs [@RicHaRDsON1996; @Richardson1995] since the PrEsence Of cycLes poses a numbEr of additiONAL challenGes aND iNTroduces additiOnal nOn-identifiABility. CoNsequEntly theRe has been CONsiderabLy lEss WorK foCUSiNg on directed gRAPhs wItH feedbaCk bOth in teRms Of iDenTifIaBility assUmptions AnD sEaRcH alGoritHMs. [@SpirteS1995] dIscUsSed The CMc, And [@RicHardsOn1996; @RiChArDSon1995] DiscussED tHE cFC fOr dCg modEls AnD intrOducED thE polynoMial-time cYclIC cauSaL dIscoverY (CCD) algorithm [@riChardson1996] foR rEcoVering THE Markov eQuivalence class for DCGs. RECently, [@cLaaSsen2013lEarnIng] introdUceD the FCi$+$ alGOrithm For recOveriNg The mARkov eQUIvAleNcE class for sPARse dCGs, wHiCh alSo assumEs the CFC. As with DAG mODelS, the CFC for cycLic ModeLS Is ExtREmELy rEsTRicTIVe since it is more RestrictivE tHAn The CFC for Dag moDeLs. In terMs of leaRning ALgorithMs that do nOt require ThE CFC, ADDitIonal assumPtions arE typicallY RequiREd. for exAmpLe [@mooiJ2011cAusAl] proVed ideNTifIabilIty for BiVariatE GausSiAn cyclic Graphical models with addiTive noIse whIch Does not reQuiRE thE CFC while Many Approaches HavE beEn stuDieD For leArniNG

imality assumptions have b een introd uced. Th ese c ondi tion s are weaker t h an t he CFC but do not guar antee m o deli de ntifi ability [@ Z h ang 20 13 ].On th e oth erhand, t he recentl y i nt roduced spar s es t Markov r epr esentation ( SMR ) andfr uga l ity a ssu mptio ns [@f o rster2 015frugal ;@ Raskut t i2013;@ v an 2013 ell] provide an a l te r native that is milde rt ha n the CF C and is s uf ficie n t to en s ur e i den t ifiability. T he main dow n sid e of t he [S M R]{} a nd fr ug a lit y assumptio ns r elative t o theC FC is t h at the[SMR]{ } a ndfrug a li ty as su m pti o ns ar e su fficient c on ditio ns f o r m odel id enti fiabi lity only whe n e xhau s tiv e sea rches ove rthe D AG spa ce ar epossible [@Rask utti 2013], wh ile t heCF C iss uffici ent fo r polyn omial-t i meal g o r it hms [@Glymour1987; @ S p ir tes1991; @Spir t es 20 0 0] for l ea rni ng e q u ivale ncec la ss of sp arse g r ap hs . Whil ethe DA Gfra mew ork i s use ful in many ap plica t ions, it is li m ited since fe e db a c kl oops ar e known toofte n exi st ( s ee e. g ., [@ Richa rd s on 1 996; @Richardson199 5] ). Hen ce, d irected graph s with dir e c t ed cycle s [@ S pi r tes2000] are m ore a ppropriate to model such feedbac k. Howeve r learning di rec ted cy c l ic graphical (D C G ) mo de ls from da ta is c ons ide rab lymo re challe nging th an l ea rn ing DAGm odels [@ Ri cha rd son 1996; @Richa rdson 1995 ]si n cethe pre s en c e ofcy cl es p ose sa num bero f a ddition al challe nge s and i nt roduces additional n on -identifia bi lit y. Con s e quentlythere has been consider a bly les s w ork f ocus ing on di rec ted gr aph s withfeedba ck bo th in t ermso f i den ti fiabilitya s sum ption sandsearchalgorithms. [@Spir t es1 995] discusse d t he C M C ,and [@ R ich ar d son 1 9 96; @Richardson 1995] disc us s ed the CFC f o r D CG models and in trodu c ed thepolynomia l-time cy cl ic c a u sal discovery (CCD) a lgorithm[ @Rich a rd son19 96] for r ec ove ringthe Ma r kov equi valenc eclassfor D CG s. Recen tly, [@claassen2013lear ning]intro duc ed the FC I$+ $ al gorithm f or r ecoveringthe Ma rkovequ i valen ce c l as s f o r spa rseD CGs, whic h a lso a ss umes the CF C . AswithDAG models , th e CFC for cyclicm odels is extre mely r est ric t ivesi nce it is more re st r i ctive th an the CFC fo r DAG mo de l s. In terms of le arninga l go r ithmsthat do not requ ire t h e CFC,ad di t ionalassu mp tionsare ty p ical l y required. For e xampl e [@moo i j20 11cau sa l] prov e d id entifiabil ity for biv ariate Gau ssian cyclic g raphic almo dels witha dditive n oisewhich d oe s no t r equire the C FC wh ilema nyapproache s ha v ebe e n s tudi ed fo rlear ning

imality assumptions_have been_introduced. These conditions are_weaker than_the_CFC but_do_not guarantee model_identifiability [@Zhang2013]. On the_other hand, the recently_introduced sparsest Markov_representation_(SMR) and frugality assumptions [@forster2015frugal; @Raskutti2013; @van2013ell] provide an alternative that is milder than the_CFC_and is_sufficient_to_ensure identifiability. The main downside_of the [SMR]{} and frugality assumptions_relative to_the CFC is that the [SMR]{} and frugality assumptions_are_sufficient conditions for_model identifiability only when exhaustive searches over the DAG_space are possible [@Raskutti2013], while the CFC_is sufficient for_polynomial-time_algorithms [@Glymour1987;_@Spirtes1991; @Spirtes2000] for learning_equivalence class of sparse graphs. While the_DAG framework is useful in many_applications, it is limited since feedback loops_are known to often exist (see_e.g., [@Richardson1996; @Richardson1995]). Hence, directed graphs_with directed_cycles [@Spirtes2000] are more appropriate to_model such feedback._However learning_directed cyclic graphical_(DCG) models from data is considerably_more challenging than_learning DAG models [@Richardson1996; @Richardson1995] since the_presence_of cycles poses_a_number_of additional_challenges and introduces_additional_non-identifiability. Consequently_there_has been considerably less work focusing_on_directed graphs with feedback both in terms_of identifiability assumptions and_search_algorithms. [@Spirtes1995] discussed the_CMC, and [@Richardson1996; @Richardson1995] discussed the_CFC for DCG models and introduced_the polynomial-time_cyclic causal_discovery (CCD) algorithm [@Richardson1996] for recovering the Markov equivalence class for DCGs._Recently, [@claassen2013learning] introduced the FCI$+$ algorithm for_recovering the Markov equivalence_class for_sparse_DCGs, which also_assumes_the CFC._As with DAG models, the CFC for_cyclic models_is extremely restrictive since it is_more restrictive than the_CFC_for DAG models. In terms of_learning algorithms that do not require_the CFC, additional assumptions are_typically_required._For example [@mooij2011causal] proved identifiability for_bivariate Gaussian cyclic graphical models with_additive noise which_does not require the CFC while many_approaches_have been studied for learning

–title($sim\_contextTitle$), and context–introduction ($sim\_contextConcl$). We turn to the context-based features. First, we focus on the features that consider the relation between the citation and the citation context: 1. Context-based features 1. contextMeta\_authorMentioned 2. contextMeta\_appearAlone 3. contextMeta\_appearFirst In this group, $contextMeta\_authorMentioned$ has the highest correlation coefficient. This feature indicates whether the names of the authors appear in the citation context (e.g., “Smith et al. \[4\]”). The second group of context-based features are based on the meaning of the words in the citation context: 1. Context-based features 1. contextLex\_relevant 2. contextLex\_recent 3. contextLex\_extreme 4. contextLex\_comparative 5. contextLexOsg\_wnPotency 6. contextLexOsg\_wnEvaluative 7. contextLexOsg\_wnActivity 8. contextLexOsg\_giPotency 9. contextLexOsg\_giEvaluative 10. contextLexOsg\_giActivity 11. contextLexEmo\_emo 12. contextLexEmo\_polarity We used several different types of lexicons to capture different aspects of semantics in the citation contexts, including sentiment, emotion, and  semantic differential categories. As we mentioned in the preceding section, $contextLexEmo\_polarity$ is the number of words in the citation context that are labeled either positive or negative and $contextLexEmo\_emo$ is the number of words that are labeled with any of the eight basic emotions. Our hope was that any kind of sentiment or emotion in the words in the citation context might indicate that the citation is influential, even if the sentiment or emotion is negative, but Figure \[fig:correlation\] shows that neither feature has a high correlation with the gold labels. This suggests to us that it might be better to split these features into more specific features for each of the possible categories. To test this idea, we split $contextLexEmo\_polarity$ into two features, one for positive polarity and one for negative

– title($sim\_contextTitle$), and context – introduction ($ sim\_contextConcl$). We turn to the context - based features. foremost, we concentrate on the features that consider the relation back between the quotation and the citation context: 1. Context - based features 1. contextMeta\_authorMentioned 2. contextMeta\_appearAlone 3. contextMeta\_appearFirst In this group, $ contextMeta\_authorMentioned$ have the highest correlation coefficient. This feature indicates whether the names of the generator appear in the citation context (for example, “ Smith et al. \[4\ ] ”). The second group of context - based features are based on the meaning of the words in the citation context: 1. Context - based features 1. contextLex\_relevant 2. contextLex\_recent 3. contextLex\_extreme 4. contextLex\_comparative 5. contextLexOsg\_wnPotency 6. contextLexOsg\_wnEvaluative 7. contextLexOsg\_wnActivity 8. contextLexOsg\_giPotency 9. contextLexOsg\_giEvaluative 10. contextLexOsg\_giActivity 11. contextLexEmo\_emo 12. contextLexEmo\_polarity We use several different types of dictionary to capture different aspects of semantics in the citation context, including sentiment, emotion, and   semantic differential categories. As we mentioned in the preceding incision, $ contextLexEmo\_polarity$ is the number of words in the citation context that are labeled either positive or negative and $ contextLexEmo\_emo$ is the number of words that are labeled with any of the eight basic emotions. Our hope was that any kind of sentiment or emotion in the words in the citation context might indicate that the citation is influential, even if the sentiment or emotion is damaging, but Figure   \[fig: correlation\ ] shows that neither feature of speech have a gamey correlation with the gold label. This suggests to us that it might be better to split these features into more specific feature for each of the possible categories. To test this idea, we split $ contextLexEmo\_polarity$ into two feature, one for positive polarity and one for negative

–titpe($sim\_contextTitle$), and coktext–introductiou ($sim\_coitextCohcl$). We tufn to the context-based featuces. Dirst, we focus on the featufes that bonsider rhe celation between the citation zkd thz ritation contexj: 1. Context-baved features 1. eontextMeta\_authorMentioned 2. contqxtMeta\_spoearAlone 3. conuexeMetz\_appearFirst In this group, $contextMsta\_authmrMentioned$ hss the highest correlation coevficient. This featkre indicatga wrwther the naoes of the authors appgar in the citation context (e.g., “Soith zt al. \[4\]”). The sgeind croup of coitext-bwsed features are baved on yhe meaning of thx woeds in the citation cmntext: 1. Context-basgd featurev 1. contextLex\_reoecant 2. congwxtUex\_derenf 3. coitextLex\_extdeme 4. cintextLex\_comparativt 5. contextLexKsg\_wnPjtqncy 6. contextLexOsg\_wnEvaluative 7. ckntextLexOsg\_wnActivity 8. contextLexOsg\_giPltency 9. contextLexOsg\_giEvaluative 10. contextLexOsg\_giAwtivivy 11. gontdztPexEmo\_emo 12. contextLexEmo\_polarity We used sedsrsl different typcs of lexicons to vaotite different arpects of semantics in the fitatiog conrexts, incjudimg sentiment, emotion, and  senantic diffegentual categories. As wz mentioned nn the precrding section, $contextLerEmo\_pomarity$ is tje number uf words in the zitsthon conttbt that are labelqd either posntive or negstive wnd $contexhLexEmo\_emo$ is the number of wlrgs that arf labeled with any of the eight basic emotionx. Mur hope waf thab any kind of sqntiment or emption iu the dords in tge citavion context might indicadg that the civation is infouenrial, evdv if the sentikent or emotion is begative, but Figurc \[fig:cudrelation\] shows thqt neither featire haf w iigh smrrelation whth ghe bold uabels. Ukis ruggrsts to us that it mhght be better to spliy bhese feajures intj more specifoc features for eafh of thx possoblg categories. To test this idea, ws split $clntcxtLexEmo\_polawity$ into two fectures, one for positive polarity and one for negative

–title($sim\_contextTitle$), and context–introduction ($sim\_contextConcl$). We turn to features. we focus the features that citation the citation context: Context-based features 1. 2. contextMeta\_appearAlone 3. contextMeta\_appearFirst In this $contextMeta\_authorMentioned$ has the highest correlation coefficient. This feature indicates whether the names of authors appear in the citation context (e.g., “Smith et al. \[4\]”). The second of features based the meaning of the words in the citation context: 1. Context-based features 1. contextLex\_relevant 2. contextLex\_recent contextLex\_extreme 4. contextLex\_comparative 5. contextLexOsg\_wnPotency 6. contextLexOsg\_wnEvaluative 7. 8. contextLexOsg\_giPotency 9. contextLexOsg\_giEvaluative contextLexOsg\_giActivity 11. contextLexEmo\_emo 12. contextLexEmo\_polarity used different types lexicons capture aspects of semantics the citation contexts, including sentiment, emotion, and semantic differential categories. As we mentioned in the preceding section, is the words in citation that labeled either positive and $contextLexEmo\_emo$ is the number of labeled with any of the eight basic emotions. hope was any kind of sentiment or emotion the words in the citation context might indicate the citation is influential, even if the sentiment or emotion is negative, but Figure \[fig:correlation\] neither feature has a correlation with the labels. suggests us it might better to split these features into more specific features for each the possible categories. To test this idea, we split $contextLexEmo\_polarity$ features, for positive polarity one for negative

–title($sim\_contextTitle$), and coNtext–introDuctiOn ($sIm\_cOnTextconcL$). We turn to the coNText-Based features. First, we foCus on ThE FeatUReS that ConsideR ThE RElaTiOn BetWeEN tHe citAtiOn and thE citation cOntExT: 1. Context-baseD FeAtures 1. contExtmeta\_authorMeNtiOned 2. coNtExtmEta\_apPeaRAlonE 3. conteXTMeta\_aPpearFirsT IN This grOUp, $conteXTmeTa\_auThorMentioned$ has tHE hIGhest correlatiOn coefFiCIeNT. thiS feAture indicAtEs wheTHer the nAMeS OF The AUthors appear iN the citatioN ConText (e.g., “smIth ET al. \[4\]”). The SeconD gROup Of context-baSed fEatures arE based ON the meaNIng of thE words In tHe cItatIOn CoNteXt: 1. cOntEXt-BasED feAtures 1. coNtExTLex\_rElevANT 2. COnteXtLEx\_reCent 3. cOntextLex\_extrEme 4. ContEXtLEx\_comParatIve 5. cOnTextLExOsg\_wNPoteNcY 6. contextLexOsg\_wNEvaLuative 7. coNteXtlexosG\_wnAcTIvity 8. cOntExtlexOsg\_gIPotencY 9. ConTeXTlExosg\_giEvaluative 10. conTeXTleXOsg\_giAcTivity 11. COnTeXTLexEmo\_eMo 12. ConTextlEXEmo\_pOlarITy we used seVeral dIFfErEnt typeS oF lexicOnS to CapTure dIFferEnt aspEcts of seMantiCS in the citation COntexts, includINg SENtIMent, EmoTion, and  semaNtic DIffeRentIAl CatEGorieS. As we MeNTiONed in the preceding seCtIon, $conTextLExEmo\_polarity$ Is the numbeR OF Words in tHe ciTAtIOn context that aRe labEled either POsitive oR negaTive and $cOntextLexeMO\_emo$ is thE nuMbeR of WorDS ThAt are labeled wITH any Of The eighT baSic emotIonS. OuR hoPe wAs That any kiNd of sentImEnT oR eMotIon in THe words iN tHe cItAtiOn conTExt migHt indIcatE tHaT The CitatioN Is INFlueNtIaL, eveN if ThE sentImenT Or eMotion iS negative, But fIgurE \[fIg:CorrelaTion\] shows that NeIther featuRe Has A high cORRelation With the gold labels. This suGGests to Us tHat it MighT be better To sPlit thEse FEatureS into mOre spEcIfiC FEaturES FoR eaCh Of the possiBLE caTegorIeS. To tEst this Idea, we split $contextlExEMo\_polarity$ intO twO feaTUReS, onE FoR PosItIVe pOLArity and one for nEgative

–title($sim\_contextTitle$ ), and con text– int rod uc tion ($s im\_contextCon c l$). We turn to the conte xt-ba se d fea t ur es. F irst, w e f o c uson t hefe a tu res t hat consid er the rel ati on between the ci tation and th e citation c ont ext: 1. C o ntext -ba sed f eature s 1. conte xt M eta\_a u thorMen t i on ed 2. contextMe t a\ _ appearAlone 3. c o nt e x tMe ta\ _appearFir st Int his gro u p, $ c ont e xtMeta\_autho rMentioned$ has the h ig hes t corre latio nc oef ficient. Th is f eature in dicate s whethe r the na mes of th e a utho r sap pea ri n t h ecit a tio n contex t(e .g.,“Smi t h e t al . \ [4\] ”). The second gr oup ofc ont ext-b asedfeat ur es ar e base d onth e meaning of th e wo rds in th e c it ati on cont e xt: 1 . Con text-ba sed fea t ure s 1. contextLex\_re le v a nt 2. cont e xt Le x \_recent 3.c ontex tLex \ _e xtreme 4. c on textLex \_ compar at ive 5. con textLe xOsg\_wn Poten c y 6. con t extLexOsg\_wn E va l u at i ve 7. contex tLex O sg\_ wnAc t iv ity 8. c on t ex t LexOsg\_giPotency 9. conte xtLexOsg\_giE valuative 10. co ntex t Le x Osg\_giActivit y 11. cont e xtLexEmo \_emo 12 . context L e xEmo\_po lar ity W e u s e dseveral diffe r e nt t yp es of l exi cons to ca ptu redif fe rent aspe cts of s em an ti cs in thec itationco nte xt s,inclu d ing se ntime nt,em ot i on, and  s e ma n t ic d if fe rent ial c atego ries . A s we me ntioned i n t h e pr ec ed ing sec tion, $contex tL exEmo\_pol ar ity $ is t h e numberof words in the citatio n contex t t hat a re l abeled ei the r posi tiv e or ne gative and$c ont e x tLexE m o \_ emo $is the num b e r o f wor ds tha t are l abeled with any of the eight basicemo tion s . O urh op e wa st hat a ny kind of sent iment or e mo t io n in the w o rds i n the c itation cont e xt migh t indicat e that th ecita t i onis influen tial, ev en if the senti m en t oremo tion i sneg ative , butF igu re \[ fig:co rr elatio n\] s ho ws thatneither feature has a h igh co rrela tio n with th e g o ldlabels. T hissuggests t o u s t hat i t m i ght b e be t te r t o spli t th e se featur e sint o mo re specific f e atu res f ore ach of the possible categor i es. To test t hisi d ea, we spli t$contextLexEmo \_p ol a r ity$ int otwo feature s, one f or posit ive po larity and on e fo r negat ive

–title($sim\_contextTitle$), and_context–introduction ($sim\_contextConcl$). We_turn to the context-based_features. First,_we_focus on_the_features that consider_the relation between_the citation and the_citation context: 1. _Context-based_features 1. contextMeta\_authorMentioned 2. contextMeta\_appearAlone __ 3.__contextMeta\_appearFirst In_this group, $contextMeta\_authorMentioned$ has the_highest correlation coefficient. This feature_indicates whether_the names of the authors appear in the_citation_context (e.g., “Smith_et al. \[4\]”). The second group of context-based features are_based on the meaning of the_words in the_citation_context: 1._ Context-based features _ 1. contextLex\_relevant _ 2. contextLex\_recent _ 3. contextLex\_extreme _ 4. contextLex\_comparative _ 5. contextLexOsg\_wnPotency _ _6. contextLexOsg\_wnEvaluative _ 7. _contextLexOsg\_wnActivity _ 8._ contextLexOsg\_giPotency 9._ contextLexOsg\_giEvaluative _ 10. contextLexOsg\_giActivity __11. contextLexEmo\_emo ___12. contextLexEmo\_polarity We_used several different_types_of lexicons_to_capture different aspects of semantics in_the_citation contexts, including sentiment, emotion, and  semantic_differential categories. As we mentioned_in_the preceding section, $contextLexEmo\_polarity$_is the number of words_in the citation context that are_labeled either_positive or_negative and $contextLexEmo\_emo$ is the number of words that are labeled_with any of the eight basic_emotions. Our hope was_that any_kind_of sentiment or_emotion_in the_words in the citation context might indicate_that the_citation is influential, even if the_sentiment or emotion is_negative,_but Figure \[fig:correlation\] shows that neither feature_has a high correlation with the_gold labels. This suggests to_us_that_it might be better to_split these features into more specific_features for each_of the possible categories. To test this idea,_we_split $contextLexEmo\_polarity$ into two features, one_for_positive polarity and one for negative

one-pion exchange (see the low-energy part of the $c_{1}$ coefficient given in Fig. \[cap:coeffs\]). This feature, which apparently contradicts the statement often made in the past that this contribution is absent in $P_{\gamma}^{(th)}$, is due to the incorporation in our work of the spin term in Eq. (\[eq:Siegert E1\]), which represents a higher order term in $q$. This correction also explains the difference in the behavior of the $c_{1}$ coefficient with the Oka’s result [@Oka:1983sp]. We note that, because the $M1$ transition dominates at the threshold and we only use the impulse approximation for its matrix element, there should be approximately a $-5\%$ correction to $A_{\gamma}^{(th)}$ (also $P_{\gamma}^{(th)}$) when two-body effects are included in $F_{M1}$. On the other hand, as $\tilde{F}_{E1_{5}}$ is calculated using the Siegert theorem, it should be reliable up to the correction of $\tilde{F}_{E1_{5}}^{(S')}$ from the PNC exchange charge at $O(1)$. When the photon energy gets larger, one can see immediately that the asymmetry gets smaller. A prediction using the DDH best values is shown in Fig. \[cap:Ag DDH\]. In this figure, as soon as the photon energy reaches 1 MeV above the threshold, the asymmetry drops by one order of magnitude. Moreover, the sign changes around $\omega_{\gamma}=4$ MeV. This implies that a higher sensitivity ($\sim10^{-9}$) is needed for any experiment targeting at the kinematic range away from the threshold. Our calculation is consistent with the work by Khriplovich and Korkin [@Khriplovich:2000mb], but is widely different from the one by Oka [@Oka:1983sp]. In the following, we make a closer comparison with these works and, then, present results for the contribution of various PNC two-body currents considered for the first time. Comparison with Oka’s work \[sub:CompOka\] ------------------------------------------ The major difference comes from the pion sector. In Ref. [@Khriplovich:2000mb], where the scattering wave functions are obtained from the zero-range approximation and the deuteron is purely a $^{3}S_{1}$ state, a simple angular momentum consideration leads to

one - pion exchange (see the low - energy character of the $ c_{1}$ coefficient render in Fig. \[cap: coeffs\ ]). This feature, which apparently contradicts the affirmation often made in the past that this contribution is lacking in $ P_{\gamma}^{(th)}$, is due to the incorporation in our work of the tailspin condition in Eq. (\[eq: Siegert E1\ ]), which represents a high order term in $ q$. This discipline also explains the difference in the behavior of the $ c_{1}$ coefficient with the Oka ’s result [ @Oka:1983sp ]. We note that, because the $ M1 $ transition dominates at the doorway and we only use the impulse approximation for its matrix chemical element, there should be approximately a $ -5\%$ correction to $ A_{\gamma}^{(th)}$ (also $ P_{\gamma}^{(th)}$) when two - body consequence are included in $ F_{M1}$. On the other hand, as $ \tilde{F}_{E1_{5}}$ is calculated using the Siegert theorem, it should be dependable up to the correction of $ \tilde{F}_{E1_{5}}^{(S')}$ from the PNC exchange charge at $ O(1)$. When the photon energy gets larger, one can see immediately that the asymmetry gets smaller. A prediction using the DDH best values is shown in Fig. \[cap: Ag DDH\ ]. In this figure, as soon as the photon energy reaches 1 MeV above the doorway, the asymmetry drops by one order of order of magnitude. furthermore, the sign changes around $ \omega_{\gamma}=4 $ MeV. This implies that a higher sensitivity ($ \sim10^{-9}$) is need for any experiment targeting at the kinematic range away from the threshold. Our calculation is reproducible with the work by Khriplovich and Korkin [ @Khriplovich:2000 mb ], but is widely different from the one by Oka [ @Oka:1983sp ]. In the following, we make a closer comparison with these work and, then, present results for the contribution of various PNC two - body currents considered for the first clock time. Comparison with Oka ’s work \[sub: CompOka\ ] ------------------------------------------ The major difference comes from the pion sector. In Ref. [ @Khriplovich:2000 mb ], where the scattering wave functions are prevail from the zero - range approximation and the deuteron is purely a $ ^{3}S_{1}$ department of state, a simple angular momentum consideration leads to

onf-pion exchange (see the luw-energy part oy the $c_{1}$ coeffjcient gkven in Fig. \[cap:coeffs\]). This fxatuee, whuch apparently contradkcts the dtatemenr ofuen made in the peat that this dlntrnbntion is absent in $P_{\gamma}^{(dh)}$, is due to tve iucorporation in our work of the spin term im Fq. (\[eq:Siegert E1\]), whibh repdvstnts a higher order term in $q$. Thia correbtion also explaims the difference in the bfhavlor of the $c_{1}$ coefflcient with the Ika’s result [@Uka:1983sp]. We noue that, becaose the $M1$ transition dominates ag the threshold abd ag only use tie impllse approximation fos its mstrix element, bhere shiuld be approximately a $-5\%$ correction to $W_{\gamma}^{(th)}$ (dlao $P_{\gamma}^{(th)}$) when rwi-body effacts qre indlndes in $F_{L1}$. Oi the other hand, as $\tiode{F}_{E1_{5}}$ is calculated ufpmg the Siegedt thejrqm, it should be reliable up to the corrtctioh of $\tilde{F}_{E1_{5}}^{(S')}$ from the PNC exchange charge aj $O(1)$. When thq photon energy gets larger, one can see immediateny thet tht qsymmdrrj gets smaller. A prediction using the DDH bese vslles is shown in Flg. \[cap:Ag DDH\]. In thos fodure, as soon xs the phkton energy reached 1 MeV wbove the threfholc, the asymmetry drops by onw order of mcgnutude. Moreover, the sign changzs aroond $\omrga_{\gamma}=4$ MeV. This implizs thaf a higher densitivifh ($\sim10^{-9}$) is needed wor aty experiment targeting at the kinenatie range xway from ehe threshlld. Oms calculation is clnsisjent whth the wogk by Khriplovich and Korkin [@Khcmplovich:2000mb], buj iv wpdely difyerent from the one by Oka [@Oka:1983sp]. Nn the fjllowkng, we makv a closec comparison with these wmtks and, then, 'resent rqsulrs fir the zuntribution of various INE two-body currents considercd fot fhe first time. Cimpqrison with Oka’x wurk \[slb:Cpm[Mka\] ------------------------------------------ The major difwerdmce cumes from tme oion sector. In Ref. [@Khripnovidh:2000mb], where the scaytcring wavg functiogs are obtainrd from the zero-rajge a'proxioatiom agd the deuteron is purely a $^{3}S_{1}$ atate, a slmpje angular mjmenbum sonsideratnon leads to

one-pion exchange (see the low-energy part of coefficient in Fig. This feature, which made the past that contribution is absent $P_{\gamma}^{(th)}$, is due to the incorporation our work of the spin term in Eq. (\[eq:Siegert E1\]), which represents a order term in $q$. This correction also explains the difference in the behavior the coefficient the result [@Oka:1983sp]. We note that, because the $M1$ transition dominates at the threshold and we only the impulse approximation for its matrix element, there be approximately a $-5\%$ to $A_{\gamma}^{(th)}$ (also $P_{\gamma}^{(th)}$) when effects included in On other as $\tilde{F}_{E1_{5}}$ is using the Siegert theorem, it should be reliable up to the correction of $\tilde{F}_{E1_{5}}^{(S')}$ from the PNC charge at the photon gets one see immediately that gets smaller. A prediction using the is shown in Fig. \[cap:Ag DDH\]. In this as soon the photon energy reaches 1 MeV the threshold, the asymmetry drops by one order magnitude. Moreover, the sign changes around $\omega_{\gamma}=4$ MeV. This implies that a higher sensitivity ($\sim10^{-9}$) for any experiment targeting the kinematic range from threshold. calculation consistent with work by Khriplovich and Korkin [@Khriplovich:2000mb], but is widely different from one by Oka [@Oka:1983sp]. In the following, we make a with works and, then, results for the contribution various two-body currents considered for time. with ------------------------------------------ major comes from the pion In Ref. [@Khriplovich:2000mb], where the wave functions are obtained the deuteron is purely a $^{3}S_{1}$ state, a angular momentum consideration leads to

one-pion exchange (see the low-eNergy part oF the $c_{1}$ CoeFfiCiEnt gIven In Fig. \[cap:coeffs\]). tHis fEature, which apparently cOntraDiCTs thE StAtemeNt often MAdE IN thE pAsT thAt THiS contRibUtion is Absent in $P_{\gAmmA}^{(tH)}$, is due to the iNCoRporation iN ouR work of the spIn tErm in EQ. (\[eQ:SiEGert E1\]), WhiCh repResentS A higheR order terM iN $Q$. This cORrectioN ALsO expLains the differencE In THe behavior of thE $c_{1}$ coefFiCIeNT WitH thE Oka’s resulT [@OKa:1983sp]. WE Note thaT, BeCAUSe tHE $M1$ transition dOminates at tHE thResholD aNd wE Only usE the iMpULse ApproximatiOn foR its matriX elemeNT, there sHOuld be aPproxiMatEly A $-5\%$ corREcTiOn tO $A_{\GAmmA}^{(Th)}$ (AlsO $p_{\gaMma}^{(th)}$) wheN tWo-Body eFfecTS ARE incLudEd in $f_{M1}$. On tHe other hand, as $\TilDe{F}_{E1_{5}}$ IS caLculaTed usIng tHe siegeRt theoRem, it ShOuld be reliable uP to tHe correctIon Of $\TilDe{f}_{E1_{5}}^{(S')}$ frOM the PNc exChaNge charGe at $O(1)$. WhEN thE pHOTOn Energy gets larger, onE cAN SeE immediaTely thAT tHe ASymmetry GeTs sMallER. a predIctiON uSing the DdH best VAlUeS is showN iN Fig. \[caP:AG DDh\]. In This fIGure, As soon As the phoTon enERgy reaches 1 MeV aBOve the threshoLD, tHE AsYMmetRy dRops by one orDer oF MagnItudE. moReoVEr, the Sign cHaNGeS Around $\omega_{\gamma}=4$ MeV. thIs implIes thAt a higher sensItivity ($\sim10^{-9}$) IS NEeded for Any eXPeRIment targeting At the Kinematic rANge away fRom thE threshoLd. Our calcULAtion is cOnsIstEnt WitH THe Work by KhriploVICh anD KOrkin [@KhRipLovich:2000mB], buT is WidEly DiFferent frOm the one By okA [@OKa:1983Sp]. IN the fOLlowing, wE mAke A cLosEr comPArison With tHese WoRkS And, Then, preSEnT REsulTs FoR the ConTrIbutiOn of VAriOus PNC tWo-body curRenTS conSiDeRed for tHe first time. CoMpArison with okA’s wOrk \[sub:cOMpOka\] ------------------------------------------ The Major difference comes froM The pion SecTor. In ref. [@KHriplovicH:2000mb], Where tHe sCAtteriNg wave FunctIoNs aRE ObtaiNED fRom ThE zero-range APProXimatIoN and The deutEron is purely a $^{3}S_{1}$ statE, A siMple angular moMenTum cONSiDerATiON leAdS To

one-pion exchange (see th e low-ener gy pa rtofth e $c _{1} $ coefficientg iven in Fig. \[cap:coeffs\ ]). T hi s fea t ur e, wh ich app a re n t lyco nt rad ic t sthe s tat ement o ften madeinth e past thatt hi s contribu tio n is absentin$P_{\g am ma} ^ {(th) }$, is d ue tot he inc orporatio ni n ourw ork oft h espin term in Eq. (\[e q :S i egert E1\]), w hich r ep r es e n tsa h igher orde rtermi n $q$.T hi s c orr e ction also ex plains thed iff erence i n t h e beha viorof the $c_{1}$ co effi cient wit h theO ka’s re s ult [@O ka:198 3sp ]. Wen ot etha t, bec a us e t h e $ M1$ tran si ti on do mina t e s at t hethre shold and we onlyuse the imp ulseappro xima ti on fo r itsmatri xelement, thereshou ld be app rox im ate ly a $- 5 \%$ co rre cti on to $ A_{\gam m a}^ {( t h ) }$ (also $P_{\gamma} ^{ ( t h) }$) when two-b o dy e f fects ar einc lude d in $F _{M1 } $. On theotherh an d, as $\t il de{F}_ {E 1_{ 5}} $ isc alcu latedusing th e Sie g ert theorem, i t should be re l ia b l eu p to th e correctio n of $\ti lde{ F }_ {E1 _ {5}}^ {(S') }$ fr o m the PNC exchangech arge a t $O( 1)$. When th e photon e n e r gy getslarg e r, one can see im media tely thatt he asymm etrygets sma ller. A p r e dictionusi ngthe DD H be st values iss h ownin Fig. \ [ca p:Ag DD H\] . I n t his f igure, as soon as t he p ho ton ener g y reache s1 M eV ab ove t h e thre shold , th eas y mme try dro p sb y one o rd er o f m ag nitud e. M o reo ver, th e sign ch ang e s ar ou nd $\omeg a_{\gamma}=4$ M eV. This i mp lie s that a highersensitivity ($\sim10^{- 9 }$) isnee ded f or a ny experi men t targ eti n g at t he kin emati cran g e away f ro m t he threshold . Our calc ul atio n is co nsistent with thew ork by Khriplovi chandK o rk in[ @K h rip lo v ich : 2 000mb], but iswidely dif fe r en t from the one b y Oka [ @Oka:19 83sp] . In the followin g, we mak ea cl o s ercomparison with th ese works and,t he n, pr ese nt res ul tsfor t he con t rib ution of va ri ous PN C two -b ody curr ents considered for the first time . Compariso n w i thOka’s wor k \[ sub:CompOk a\] -- ----- --- - ----- ---- - -- --- - ----- ---- - ----- Th e m ajo r di fference co m e s fr om th e p i on sec tor. In Ref. [@Khripl o vich:2000mb],wher e the sc a tter in g wave functio nsar e obtained f rom the zer o-rangeap p roxim ationand th e deute r o ni s pure ly a $^ {3}S_{1}$ st at e , a sim pl ea ngular mom en tum co nsider a tion l eads to

one-pion_exchange (see_the low-energy part of_the $c_{1}$_coefficient_given in_Fig._\[cap:coeffs\]). This feature,_which apparently contradicts_the statement often made_in the past_that_this contribution is absent in $P_{\gamma}^{(th)}$, is due to the incorporation in our work_of_the spin_term_in_Eq. (\[eq:Siegert E1\]), which represents_a higher order term in_$q$. This_correction also explains the difference in the behavior_of_the $c_{1}$ coefficient_with the Oka’s result [@Oka:1983sp]. We note that, because the_$M1$ transition dominates at the threshold_and we only_use_the_impulse approximation for its_matrix element, there should be approximately_a $-5\%$ correction to $A_{\gamma}^{(th)}$ (also_$P_{\gamma}^{(th)}$) when two-body effects are included in_$F_{M1}$. On the other hand, as_$\tilde{F}_{E1_{5}}$ is calculated using the_Siegert theorem,_it should be reliable up_to the correction_of $\tilde{F}_{E1_{5}}^{(S')}$_from the PNC_exchange charge at $O(1)$. When the photon_energy gets larger,_one can see immediately that the_asymmetry_gets smaller. A_prediction_using_the DDH_best values is_shown_in Fig._\[cap:Ag_DDH\]. In this figure, as soon_as_the photon energy reaches 1 MeV above_the threshold, the asymmetry_drops_by one order of_magnitude. Moreover, the sign changes_around $\omega_{\gamma}=4$ MeV. This implies that_a higher_sensitivity ($\sim10^{-9}$)_is needed for any experiment targeting at the kinematic range away_from the threshold. Our calculation is_consistent with the work_by Khriplovich_and_Korkin [@Khriplovich:2000mb], but_is_widely different_from the one by Oka [@Oka:1983sp]. In_the following,_we make a closer comparison with_these works and, then,_present_results for the contribution of various_PNC two-body currents considered for the_first time. Comparison with Oka’s work_\[sub:CompOka\] ------------------------------------------ The_major_difference comes from the pion_sector. In Ref. [@Khriplovich:2000mb], where the_scattering wave functions_are obtained from the zero-range approximation and_the_deuteron is purely a $^{3}S_{1}$ state,_a_simple angular momentum consideration leads to

A quick and revealing method of looking at global changes in the system is an energy track. This is illustrated by  \[enerfs\] which shows the energy statistics for one of our mergers. Energy and time are normalised by setting the total mass, $M_{tot}$, the outer radius of the system, $R_{max}$ and $G$ equal to unity. The thermal energy is displayed multiplied by a factor of 8, as the total gas mass is ${1 \over 8}$ that of the dark matter. The effect of the merger is clearly visible on  \[enerfs\], however it is impossible to judge the accuracy of the energy conservation due to the scaling.  \[enercons\] expands the system total energy and scales it to the initial value. There are nearly 1000 timesteps in this plot with typical energy conservation better than 1 part in 10000 per step. The same accuracy, smoothing and timestepping parameters were used in all of the runs detailed in this paper. The combined code was tested by observing the evolution of a static configuration. A single isolated cluster was set up following the prescription given above. This consisted of 8192 gas and 8192 dark matter particles in what should be a nearly stable arrangement. For this run the total energy is conserved to better the 0.4 percent and the start and finish parameters are listed in Table 1. Some expansion at the edge is to be expected as here the gas is not in pressure balance. As in our previous paper () the choice of centre is very important when plotting logarithmic density profiles. For the static run we tried the ‘true’ centre (the position about which the initial profile was built), the position of the gas particle with the highest local density, the mean position of the 100 or so densest gas particles, the most tightly bound dark matter particle and the mean of the 100 or so most tightly bound particles. All five of these positions remained close together, within less than the gravitational softening length. In what follows the mean position of the 100 or so densest gas particles is used as the expansion centre for profiling and radial measurements.  \[densprof\] shows the initial and final density profiles for the gas and the dark matter, with the end-state taken after approximately the same length of time as the collisions discussed below. The expansion of the edge can readily be seen, as can a very small increase in the core size of both the gas and the dark matter. Some rearrangement in the centre is almost certain to occur as

A quick and revealing method of looking at ball-shaped change in the system is an department of energy track. This is illustrate by   \[enerfs\ ] which shows the department of energy statistic for one of our mergers. Energy and time are normalised by setting the full mass, $ M_{tot}$, the outer spoke of the system, $ R_{max}$ and $ G$ equal to unity. The thermal energy is displayed multiplied by a factor of 8, as the entire gas mass is $ { 1 \over 8}$ that of the dark topic. The effect of the merger is intelligibly visible on   \[enerfs\ ], however it is impossible to judge the accuracy of the department of energy conservation due to the scaling.   \[enercons\ ] expands the organization total energy and scales it to the initial value. There be nearly 1000 timesteps in this plot with typical energy conservation better than 1 part in 10000 per dance step. The same accuracy, smoothing and timestepping parameters were used in all of the runs detailed in this paper. The combined code was tested by observing the evolution of a static shape. A individual isolated bunch was rig up follow the prescription given above. This consist of 8192 gas and 8192 dark matter particles in what should be a about stable arrangement. For this run the total energy is conserved to better the 0.4 percentage and the start and finish parameters are listed in Table   1. Some expansion at the edge is to be expected as here the gas is not in press balance. As in our previous paper () the choice of centre is very authoritative when plotting logarithmic density profiles. For the inactive run we tried the ‘ true ’ centre (the situation about which the initial profile was build), the military position of the gas atom with the highest local density, the mean position of the 100 or indeed densest gas particles, the most tightly bound colored matter particle and the mean of the 100 or so most tightly bound particles. All five of these positions remained close in concert, within less than the gravitational softening length. In what comply the mean position of the 100 or so densest gas particles is used as the expansion center for profiling and radial measurements.   \[densprof\ ] shows the initial and final density profiles for the gas and the black topic, with the end - state taken after approximately the same length of time as the collisions discussed downstairs. The expansion of the edge can readily be seen, as can a very little increase in the core size of both the gas and the dark matter. Some rearrangement in the center is almost certain to occur as

A qkick and revealing methoa of looking at global changss in thd system is an energy track. Vhis is iolustrated by  \[enerfs\] wfich showd the enwrgy wtatistics for one of oud meryecs. Energy and tlme are norkalised by setdivg the total mass, $M_{tot}$, the outer radits of tne system, $R_{max}$ agd $G$ qquam to unity. The thermal energy is djsplayev multiplied by a factor of 8, as the total gas mass is ${1 \over 8}$ thwt of the dqrk iqtter. The efwect of tht kerger is dlearly visible on  \[enerfs\], howevdr it is impossuboe hm judge the accugacy of the ekvrgy cotservatoon due to the scelint.  \[enercons\] expands thx system total energi and scalas it to the initiao calue. These afw ndarmy 1000 fimestfps in this pmot with typical energy conserfaeppn better thzn 1 pawt in 10000 per step. The same accuracy, smoothitg znd timestepping paramerers were used in all of the rtns detailed in this paper. The combined code was tasted cy ibfdevlng the evolution of a static configuration. A aimgke isolated clmster was set up fplpoeyng the presctiption giben above. This condisted jf 8192 gqs and 8192 dwrk katter particles in what shiuld be a necrlt stable arrangemeut. For this xun thg totak energy is conserved tu befter the 0.4 pfrcent ans the start and fknixh paramettfs are listed in Eable 1. Somx expcnsion ag thg edge ys to be edpectcg as here the gas ls noj in psessure bapance. As in our previous paper () the choice of cantge is verv impovtant when ploteing logarithmnc densicy prowiles. For nhe statir run we triqd the ‘true’ cajtre (the posmtion abott wyich the inkgial profile wss built), nht position of the gas particke djth the highest loxal density, the mexn [odivion jx the 100 or so denresg gas oarticltf, bhe mosy tightly bound dark matfer particle and tne mean of the 100 or so most tighyly bound particled. All fite of yhefe positions remained close tofether, wihhik less than tre gvavijational soytening length. In what follows the mean 'osition of the 100 or so eensest gas particlgs ls used as tie expwnsion cettre for profiling abd radial measurekents.  \[densprof\] shows tge inidial wnd final density profiles for the gas and the dark matter, with the end-stqte taken after aplroxomateny che samq leigvh of time as the collisions discussed below. The expansimn of the edge can readily be xedn, as can a vgry small increase in the dore sizr of both the gas and the dark katter. Some rearrangemenb in the xentrs is almosv certain to occur ax

A quick and revealing method of looking changes the system an energy track. which the energy statistics one of our Energy and time are normalised by the total mass, $M_{tot}$, the outer radius of the system, $R_{max}$ and $G$ to unity. The thermal energy is displayed multiplied by a factor of 8, the gas is \over 8}$ that of the dark matter. The effect of the merger is clearly visible on however it is impossible to judge the accuracy the energy conservation due the scaling. \[enercons\] expands the total and scales to initial There are nearly timesteps in this plot with typical energy conservation better than 1 part in 10000 per step. The accuracy, smoothing parameters were in of runs detailed in The combined code was tested by of a static configuration. A single isolated cluster set up the prescription given above. This consisted 8192 gas and 8192 dark matter particles in should be a nearly stable arrangement. For this run the total energy is conserved to 0.4 percent and the and finish parameters listed Table Some at the is to be expected as here the gas is not in balance. As in our previous paper () the choice of very when plotting logarithmic profiles. For the static we the ‘true’ centre (the which initial the of gas particle with the local density, the mean position the 100 or so tightly bound dark matter particle and the mean the 100 or so most tightly bound All five of these positions remained close together, within less than the softening length. follows the mean position of the 100 or densest gas particles is as the expansion centre for profiling and radial measurements. shows initial and density profiles for gas and the matter, with the after approximately same of collisions discussed below. The expansion of edge readily be seen, as can very in the core size of both the gas and the dark matter. Some rearrangement in is almost certain to as

A quick and revealing method oF looking at GlobaL chAngEs In thE sysTem is an energy tRAck. THis is illustrated by  \[enerFs\] whiCh SHows THe EnergY statisTIcS FOr oNe Of Our MeRGeRs. EneRgy And time Are normaliSed By Setting the toTAl Mass, $M_{tot}$, thE ouTer radius of tHe sYstem, $R_{MaX}$ anD $g$ equaL to Unity. the theRMal eneRgy is dispLaYEd multIPlied by A FAcTor oF 8, as the total gas masS Is ${1 \OVer 8}$ that of the daRk mattEr. tHe EFFecT of The merger iS cLearlY Visible ON  \[eNERFs\], hOWever it is impoSsible to judGE thE accurAcY of THe enerGy conSeRVatIon due to the ScalIng.  \[enercoNs\] expaNDs the sySTem totaL energY anD scAles IT tO tHe iNiTIal VAlUe. THEre Are nearlY 1000 tImEstepS in tHIS PLot wIth TypiCal enErgy conservatIon BettER thAn 1 parT in 10000 peR steP. THe samE accurAcy, smOoThing and timestePpinG parameteRs wErE usEd In all OF the ruNs dEtaIled in tHis papeR. the CoMBINeD code was tested by obSeRVInG the evolUtion oF A sTaTIc configUrAtiOn. A sINGle isOlatED cLuster waS set up FOlLoWing the PrEscripTiOn gIveN abovE. this ConsisTed of 8192 gas And 8192 daRK matter particlES in what should BE a NEArLY staBle Arrangement. for tHIs ruN the TOtAl eNErgy iS consErVEd TO better the 0.4 percent anD tHe starT and fInish parameteRs are listeD IN table 1. SomE expANsIOn at the edge is tO be exPected as heRE the gas iS not iN pressurE balance. AS IN our prevIouS paPer () The CHOiCe of centre is vERY impOrTant wheN plOtting lOgaRitHmiC deNsIty profilEs. For the StAtIc RuN we Tried THe ‘true’ ceNtRe (tHe PosItion ABout whIch thE iniTiAl PRofIle was bUIlT), THe poSiTiOn of The GaS partIcle WIth The highEst local dEnsITy, thE mEaN positiOn of the 100 or so deNsEst gas partIcLes, The mosT TIghtly boUnd dark matter particle anD The mean Of tHe 100 or sO mosT tightly bOunD partiCleS. all fivE of theSe posItIonS REmainED ClOse ToGether, withIN LesS than ThE graVitatioNal softening length. iN whAt follows the mEan PosiTIOn Of tHE 100 oR So dEnSEst GAS particles is useD as the expaNsIOn Centre for pROfiLiNg and raDial meaSuremENts.  \[densProf\] shows The initiaL aNd fiNAL deNsity profiLes for thE gas and thE Dark mATtEr, witH thE end-stAtE taKen afTer appROxiMatelY the saMe Length Of timE aS the collIsions discussed below. The ExpansIon of The Edge can reAdiLY be Seen, as can A verY small incrEasE in The coRe sIZe of bOth tHE gAs aND the dArk mATter. Some rEArRanGEMeNt in the centRE IS alMost cErtAIn to ocCur aS

A quick and revealing me thod of lo oking at gl ob al c hang es in the syst e m is an energy track. This is i ll u stra t ed by \[enerf s \] w hic hsh ows t h eenerg y s tatisti cs for one of o ur mergers.E ne rgy and ti meare normalis edby set ti ngt he to tal mass , $M_{ t ot}$,the outer r a dius o f the sy s t em , $R _{max}$ and $G$ e q ua l to unity. The therm al en e r gyisdisplayedmu ltipl i ed by a fa c t o r o f 8, as the to tal gas mas s is ${1 \ ov er8 }$ tha t ofth e da rk matter.Theeffect of the m e rger is clearly visib leon \[e n er fs \], h o wev e riti s i mpossibl eto judg e th e a c cura cyof t he en ergy conserva tio n du e to thescali ng. \ [ener cons\] expa nd s the system to talenergy an d s ca les i t tot he ini tia l v alue. T here ar e ne ar l y 10 00 timesteps in th is p lo t with t ypical en er g y conser va tio n be t t er th an 1 pa rt in 10 000 pe r s te p. Thesa me acc ur acy , s mooth i ng a nd tim esteppin g par a meters were us e d in all of t h er u ns deta ile d in this p aper . Th e co m bi ned codewas t es t ed by observing the ev ol utionof astatic config uration. A s i ngle iso late d c l uster was setup fo llowing th e prescri ption given a bove. Thi s consiste d o f 8 192 ga s an d 8192 dark m a t terpa rticles in what s hou ldbea n ea rly stabl e arrang em en t. F orthisr un the t ot alen erg y isc onserv ed to bet te rt he0.4 per c en t andth estar t a nd fini sh p a ram eters a re listed in Tabl e1. Some e xpansion at t he edge is t obeexpect e d as here the gas is not in pres s ure bal anc e. As inour previ ous paper () the ch oice o f cen tr e i s veryi m po rta nt when plot t i nglogar it hmic densit y profiles. For th e st atic run we t rie d th e ‘t rue ’ c e ntr e( the p osition about w hich the i ni t ia l profilew asbu ilt), t he posi tiono f the g as partic le with t he hig h e stlocal dens ity, the mean pos i tiono fthe 1 00or sode nse st ga s part i cle s, th e most t ightly boun ddark mat ter particle and the me an ofthe 1 00or so mos t t i ght ly boundpart icles. All fi veof th ese posit ions re mai n ed cl oset ogether,w it hin l es s than theg r a vit ation als ofteni ng l ength. In what fo l lows the meanposi t i onoft he 1 00 or so densest ga sp a rticlesis used as th e expans io n cent re for profi ling an d ra d ial me asur eme nts.  \[ den sp r of\] sh ow st he ini tial a nd fin al den s ityp r ofiles for the g as an d the d a rkmatte r, with t h e en d-state ta ken after a pproxi mate ly th e samele ngth o f t im e as the c o llisionsdiscu ssed be lo w. T heexpans iono f theedge c anreadily b e se e n, a s ca n averysm allincreasei n the co res ize ofbo tht h e gasa nd t he dark ma tte r. So m e rearrange m enti nt he ce ntre i s almo st cert a into occuras

A quick_and revealing_method of looking at_global changes_in_the system_is_an energy track._This is illustrated_by  \[enerfs\] which shows_the energy statistics_for_one of our mergers. Energy and time are normalised by setting the total mass,_$M_{tot}$,_the outer_radius_of_the system, $R_{max}$ and $G$_equal to unity. The thermal_energy is_displayed multiplied by a factor of 8, as_the_total gas mass_is ${1 \over 8}$ that of the dark matter._The effect of the merger is_clearly visible on_ \[enerfs\],_however_it is impossible to_judge the accuracy of the energy_conservation due to the scaling.  \[enercons\]_expands the system total energy and scales_it to the initial value. There_are nearly 1000 timesteps in_this plot_with typical energy conservation better_than 1 part_in 10000_per step. The_same accuracy, smoothing and timestepping parameters_were used in_all of the runs detailed in_this_paper. The combined code_was_tested_by observing_the evolution of_a_static configuration._A_single isolated cluster was set up_following_the prescription given above. This consisted of_8192 gas and 8192_dark_matter particles in what_should be a nearly stable_arrangement. For this run the total_energy is_conserved to_better the 0.4 percent and the start and finish parameters are_listed in Table 1. Some expansion at_the edge is to_be expected_as_here the gas_is_not in_pressure balance. As in our previous paper_() the_choice of centre is very important_when plotting logarithmic density_profiles._For the static run we tried_the ‘true’ centre (the position about_which the initial profile was_built),_the_position of the gas particle_with the highest local density, the_mean position of_the 100 or so densest gas particles,_the_most tightly bound dark matter particle_and_the mean of the 100 or_so_most_tightly bound particles. All five_of these positions remained close together,_within less than the gravitational softening length. In what_follows the mean_position of the 100 or_so_densest_gas particles is used as the expansion centre for profiling_and radial_measurements.  \[densprof\] shows the_initial and final density profiles for the gas and the_dark matter, with the end-state taken after_approximately the same length of time as the collisions discussed below._The expansion of the edge can readily be_seen, as can a very small increase_in the core size_of_both the gas and the dark matter._Some_rearrangement_in_the centre is_almost certain to_occur as

study of early universe models of electroweak phase transition [@new] or inflationary scenarios along the lines of Ref. [@Barrie:2016rnv] would also be interesting. #### Acknowledgement. The work was supported in part by the Australian Research Council. Beta functions ============== For reader’s convenience here we include the relevant one-loop beta-functions, $\beta_{C}=\frac{dC}{d\ln(\mu)}$, used in our calculations: $$\begin{aligned} \beta_{g_{Y}} & = & \frac{g_{Y}^{3}}{16\pi^{2}}\frac{41}{6},\quad\beta_{g_{2}}=\frac{g_{2}^{3}}{16\pi^{2}}\left(-\frac{19}{6}\right),\quad\beta_{g_{3}}=\frac{g_{3}^{3}}{16\pi^{2}}\left(-7\right)\\ \beta_{y_{t}} & = & \frac{y_{t}}{16\pi^{2}}\left(-\frac{9}{4}g_{2}^{2}-8g_{3}^{2}-\frac{17}{12}g_{Y}^{2}+\frac{9}{2}y_{t}^{2}\right)\\ \beta_{\lambda} & = & \frac{1}{16\pi^2}\left[\lambda\left(-9g_{2}^{2}-3g_{Y}^{2}+24\lambda+12y_{t}^{2}\right)+\frac{3}{4}g_{2}^{2}g_{Y}^{2}+\frac{9}{8}g_{2}^{4}+\frac{3}{8}g_{Y}^{4}-6y_{t}^{4}+2\lambda_{h\chi}^{2}\right]\\ \beta_{\lambda_{\chi}} & = & \frac{1}{16\pi^{2}}\left(18\lambda_{\chi}+8\lambda_{h\chi}^{2}\right)\\ \beta_{\lambda_{h\chi}} & = & \frac{\lambda_{h\chi}}{16\pi^2}\left(-\frac{9}{2}g_{2}^{2}-\frac{3}{2}g_{Y}^{2}+12\lambda+6\lambda_{\chi}+8\lambda_{h\chi}+6y_{t}^{2}\right)\\ \beta

study of early universe models of electroweak phase transition [ @new ] or inflationary scenario along the line of Ref. [ @Barrie:2016rnv ] would also be interesting. # # # # Acknowledgement. The work was support in part by the Australian Research Council. Beta functions = = = = = = = = = = = = = = For lector ’s appliance here we include the relevant one - cringle beta - functions, $ \beta_{C}=\frac{dC}{d\ln(\mu)}$, used in our calculation: $ $ \begin{aligned } \beta_{g_{Y } } & = & \frac{g_{Y}^{3}}{16\pi^{2}}\frac{41}{6},\quad\beta_{g_{2}}=\frac{g_{2}^{3}}{16\pi^{2}}\left(-\frac{19}{6}\right),\quad\beta_{g_{3}}=\frac{g_{3}^{3}}{16\pi^{2}}\left(-7\right)\\ \beta_{y_{t } } & = & \frac{y_{t}}{16\pi^{2}}\left(-\frac{9}{4}g_{2}^{2}-8g_{3}^{2}-\frac{17}{12}g_{Y}^{2}+\frac{9}{2}y_{t}^{2}\right)\\ \beta_{\lambda } & = & \frac{1}{16\pi^2}\left[\lambda\left(-9g_{2}^{2}-3g_{Y}^{2}+24\lambda+12y_{t}^{2}\right)+\frac{3}{4}g_{2}^{2}g_{Y}^{2}+\frac{9}{8}g_{2}^{4}+\frac{3}{8}g_{Y}^{4}-6y_{t}^{4}+2\lambda_{h\chi}^{2}\right]\\ \beta_{\lambda_{\chi } } & = & \frac{1}{16\pi^{2}}\left(18\lambda_{\chi}+8\lambda_{h\chi}^{2}\right)\\ \beta_{\lambda_{h\chi } } & = & \frac{\lambda_{h\chi}}{16\pi^2}\left(-\frac{9}{2}g_{2}^{2}-\frac{3}{2}g_{Y}^{2}+12\lambda+6\lambda_{\chi}+8\lambda_{h\chi}+6y_{t}^{2}\right)\\ \beta

stkdy of early universe moaels of electrocwak phese trahsition [@vew] or inflationary scenariod qlong the lines of Ref. [@Barrke:2016rnv] woupd also ve iiteresting. #### Acknoxmedgemekc. The slrk caw supported in part by tve Australian Serecrch Council. Beta functions ============== For reader’f convemifnce here we igclucq ths relevant one-loop beta-functions, $\bsta_{C}=\frab{dC}{d\ln(\mu)}$, used in pur calculations: $$\begin{aligjed} \bfta_{g_{Y}} & = & \frac{g_{Y}^{3}}{16\pi^{2}}\fgac{41}{6},\quad\beta_{t_{2}}=\fras{t_{2}^{3}}{16\pi^{2}}\left(-\frac{19}{6}\rkght),\quad\beua_{y_{3}}=\frac{g_{3}^{3}}{16\pi^{2}}\lefj(-7\right)\\ \beta_{y_{t}} & = & \frac{y_{t}}{16\pi^{2}}\left(-\frac{9}{4}e_{2}^{2}-8g_{3}^{2}-\frae{17}{12}g_{Y}^{2}+\frac{9}{2}y_{t}^{2}\rithr)\\ \behd_{\lambda} & = & \hrac{1}{16\pi^2}\jeft[\lambda\lenn(-9g_{2}^{2}-3g_{Y}^{2}+24\lamtda+12y_{t}^{2}\ribht)+\frac{3}{4}g_{2}^{2}g_{Y}^{2}+\frac{9}{8}n_{2}^{4}+\frac{3}{8}j_{Y}^{4}-6y_{t}^{4}+2\oambda_{h\chi}^{2}\right]\\ \beta_{\lakbda_{\chi}} & = & \frac{1}{16\pi^{2}}\lgft(18\lambda_{\cvi}+8\mambda_{h\chi}^{2}\right)\\ \bera_{\oambdd_{h\chh}} & = & \draz{\lajbva_{h\dhi}}{16\pi^2}\lfft(-\hrac{9}{2}g_{2}^{2}-\frac{3}{2}g_{Y}^{2}+12\mambda+6\lambdq_{\chi}+8\lambda_{h\chi}+6y_{t}^{2}\rignt)\\ \fvya

study of early universe models of electroweak [@new] inflationary scenarios the lines of interesting. Acknowledgement. The work supported in part the Australian Research Council. Beta functions For reader’s convenience here we include the relevant one-loop beta-functions, $\beta_{C}=\frac{dC}{d\ln(\mu)}$, used in calculations: $$\begin{aligned} \beta_{g_{Y}} & = & \frac{g_{Y}^{3}}{16\pi^{2}}\frac{41}{6},\quad\beta_{g_{2}}=\frac{g_{2}^{3}}{16\pi^{2}}\left(-\frac{19}{6}\right),\quad\beta_{g_{3}}=\frac{g_{3}^{3}}{16\pi^{2}}\left(-7\right)\\ \beta_{y_{t}} & = & \frac{y_{t}}{16\pi^{2}}\left(-\frac{9}{4}g_{2}^{2}-8g_{3}^{2}-\frac{17}{12}g_{Y}^{2}+\frac{9}{2}y_{t}^{2}\right)\\ \beta_{\lambda} = \frac{1}{16\pi^2}\left[\lambda\left(-9g_{2}^{2}-3g_{Y}^{2}+24\lambda+12y_{t}^{2}\right)+\frac{3}{4}g_{2}^{2}g_{Y}^{2}+\frac{9}{8}g_{2}^{4}+\frac{3}{8}g_{Y}^{4}-6y_{t}^{4}+2\lambda_{h\chi}^{2}\right]\\ & & \frac{1}{16\pi^{2}}\left(18\lambda_{\chi}+8\lambda_{h\chi}^{2}\right)\\ \beta_{\lambda_{h\chi}} & = & \frac{\lambda_{h\chi}}{16\pi^2}\left(-\frac{9}{2}g_{2}^{2}-\frac{3}{2}g_{Y}^{2}+12\lambda+6\lambda_{\chi}+8\lambda_{h\chi}+6y_{t}^{2}\right)\\ \beta

study of early universe modelS of electroWeak pHasE trAnSitiOn [@neW] or inflationarY ScenArios along the lines of ReF. [@BarrIe:2016RNv] woULd Also bE intereSTiNG. #### ackNoWlEdgEmENt. the woRk wAs suppoRted in part By tHe australian ReSEaRch Council. betA functions ============== FoR reAder’s cOnVenIEnce hEre We incLude thE RelevaNt one-loop BeTA-functIOns, $\beta_{c}=\FRaC{dC}{d\Ln(\mu)}$, used in our calcULaTIons: $$\begin{alignEd} \beta_{G_{Y}} & = & \FRaC{G_{y}^{3}}{16\pi^{2}}\FraC{41}{6},\quad\beta_{g_{2}}=\FrAc{g_{2}^{3}}{16\pi^{2}}\LEft(-\frac{19}{6}\RIgHT),\QUad\BEta_{g_{3}}=\frac{g_{3}^{3}}{16\pi^{2}}\leFt(-7\right)\\ \beta_{Y_{T}} & = & \frAc{y_{t}}{16\pi^{2}}\LeFt(-\fRAc{9}{4}g_{2}^{2}-8g_{3}^{2}-\frAc{17}{12}g_{Y}^{2}+\fRaC{9}{2}Y_{t}^{2}\rIght)\\ \beta_{\lamBda} & = & \fRac{1}{16\pi^2}\left[\Lambda\LEft(-9g_{2}^{2}-3g_{Y}^{2}+24\lAMbda+12y_{t}^{2}\rIght)+\frAc{3}{4}g_{2}^{2}G_{Y}^{2}+\fRac{9}{8}g_{2}^{4}+\FRaC{3}{8}g_{y}^{4}-6y_{t}^{4}+2\LaMBda_{H\ChI}^{2}\riGHt]\\ \bEta_{\lambdA_{\cHi}} & = & \Frac{1}{16\pI^{2}}\lefT(18\LAMBda_{\cHi}+8\lAmbdA_{h\chi}^{2}\Right)\\ \beta_{\lambDa_{h\Chi}} & = & \fRAc{\lAmbda_{H\chi}}{16\pI^2}\lefT(-\fRac{9}{2}g_{2}^{2}-\fRac{3}{2}g_{Y}^{2}+12\lAmbda+6\LaMbda_{\chi}+8\lambda_{h\cHi}+6y_{t}^{2}\Right)\\ \beta

study of early universe m odels of e lectr owe akph asetran sition [@new]o r in flationary scenarios a longth e lin e sof Re f. [@Ba r ri e : 201 6r nv ] w ou l dalsobeinteres ting. ### # A ck nowledgement . The work w assupported in pa rt byth e A u stral ian Rese arch C o uncil. Beta fu nc t ions = = ======= = = == = F or reader’s conve n ie n ce here we inc lude t he re l e van t o ne-loop be ta -func t ions, $ \ be t a _ {C} = \frac{dC}{d\l n(\mu)}$, u s edin our c alc u lation s: $$ \b e gin {aligned} \ beta _{g_{Y}}& = &\ frac{g_ { Y}^{3}} {16\pi ^{2 }}\ frac { 41 }{ 6}, \q u ad\ b et a_{ g _{2 }}=\frac {g _{ 2}^{3 }}{1 6 \ p i ^{2} }\l eft( -\fra c{19}{6}\righ t), \qua d \be ta_{g _{3}} =\fr ac {g_{3 }^{3}} {16\p i^ {2}}\left(-7\ri ght) \\ \beta_ {y_ {t }}&= & \ f rac{y_ {t} }{1 6\pi^{2 }}\left ( -\f ra c { 9 }{ 4}g_{2}^{2}-8g_{3} ^{ 2 } -\ frac{17} {12}g_ { Y} ^{ 2 }+\frac{ 9} {2} y_{t } ^ {2}\r ight ) \\ \beta_{ \lambd a }&= & \fr ac {1}{16 \p i^2 }\l eft[\ l ambd a\left (-9g_{2} ^{2}- 3 g_{Y}^{2}+24\l a mbda+12y_{t}^ { 2} \ r ig h t)+\ fra c{3}{4}g_{2 }^{2 } g_{Y }^{2 } +\ fra c {9}{8 }g_{2 }^ { 4} + \frac{3}{8}g_{Y}^{4 }- 6y_{t} ^{4}+ 2\lambda_{h\c hi}^{2}\ri g h t ]\\ \bet a_{\ l am b da_{\chi}} & = & \f rac{1}{16\ p i^{2}}\l eft(1 8\lambda _{\chi}+8 \ l ambda_{h \ch i}^ {2} \ri g h t) \\ \beta_{\la m b da_{ h\ chi}} & =& \frac {\l amb da_ {h\ ch i}}{16\pi ^2}\left (- \f ra c{ 9}{ 2}g_{ 2 }^{2}-\f ra c{3 }{ 2}g _{Y}^ { 2}+12\ lambd a+6\ la mb d a_{ \chi}+8 \ la m b da_{ h\ ch i}+6 y_{ t} ^{2}\ righ t )\\ \beta

study_of early_universe models of electroweak_phase transition_[@new]_or inflationary_scenarios_along the lines_of Ref. [@Barrie:2016rnv]_would also be interesting. ####_Acknowledgement. The work was_supported_in part by the Australian Research Council. Beta functions ============== For reader’s convenience here we include the_relevant_one-loop beta-functions,_$\beta_{C}=\frac{dC}{d\ln(\mu)}$,_used_in our calculations: $$\begin{aligned} \beta_{g_{Y}} &_= & \frac{g_{Y}^{3}}{16\pi^{2}}\frac{41}{6},\quad\beta_{g_{2}}=\frac{g_{2}^{3}}{16\pi^{2}}\left(-\frac{19}{6}\right),\quad\beta_{g_{3}}=\frac{g_{3}^{3}}{16\pi^{2}}\left(-7\right)\\ \beta_{y_{t}} & =_& \frac{y_{t}}{16\pi^{2}}\left(-\frac{9}{4}g_{2}^{2}-8g_{3}^{2}-\frac{17}{12}g_{Y}^{2}+\frac{9}{2}y_{t}^{2}\right)\\ \beta_{\lambda}_& = & \frac{1}{16\pi^2}\left[\lambda\left(-9g_{2}^{2}-3g_{Y}^{2}+24\lambda+12y_{t}^{2}\right)+\frac{3}{4}g_{2}^{2}g_{Y}^{2}+\frac{9}{8}g_{2}^{4}+\frac{3}{8}g_{Y}^{4}-6y_{t}^{4}+2\lambda_{h\chi}^{2}\right]\\ \beta_{\lambda_{\chi}} & = & \frac{1}{16\pi^{2}}\left(18\lambda_{\chi}+8\lambda_{h\chi}^{2}\right)\\ \beta_{\lambda_{h\chi}}_&_= & \frac{\lambda_{h\chi}}{16\pi^2}\left(-\frac{9}{2}g_{2}^{2}-\frac{3}{2}g_{Y}^{2}+12\lambda+6\lambda_{\chi}+8\lambda_{h\chi}+6y_{t}^{2}\right)\\ \beta

book defined in Section \[sec: general order book\] represents the dynamics of the stock. Whenever the trader buys/sells $n$ units of stocks (being by an aggressive order or by the execution of a limit order), she sells/buys immediately $n$ units of the futures. Her inventory should be fully liquidated at $T$. The optimal control problem {#the-optimal-control-problem} --------------------------- We assume that the reference price $F$ of the futures is given by $$\begin{aligned} \label{eq: def F} F=\frac{P^{\mathfrak b}+P^{\mathfrak a}}{2}+S \end{aligned}$$ where $S$ is a mean-reverting process $$\begin{aligned} \label{eq: dyna S} S= S_{0}+\int_{0}^{\cdot}\rho(\hat s-S_t)dt+\int_{0}^{\cdot}\sigma(S_{t}) dW_t.\end{aligned}$$ Here, $\rho$ is the strength of mean reversion, $\hat s$ is the average of mean reversion and $\sigma:{\mathbb{R}}\mapsto {\mathbb{R}}$ is a Lipschitz bounded function representing the volatility of the process. In the following, we shall assume that $$\begin{aligned} \label{eq: ass sigma bounded support} \mbox{ the support of $\sigma$ is bounded.}\end{aligned}$$ \[rem : DS\] The above implies in particular that $S$ lies in a certain compact set $D_{S}$ as soon as $S_{0}$ does. [This could clearly be relaxed to the price of a finer analysis. ]{} The strategy of the HFT is described by the same quantities as the one of the MM in Section \[sec:marketmaker\]. The only difference lies in the fact that she constantly holds a number equal to $-I$ units of the futures $F$. We assume that buying/selling the futures leads to the payment of a proportional cost $\kappa\ge 0$. Then, the dynamics of the corresponding state process $X$ is given by $$\begin{aligned} \label{eq: dyna X HFT} X_{\tau_{i}}=\Tc_{X}(S_{\tau_{i}},P_{\tau_{i-1}},Q_{\tau_{i-1}},X_{\tau_{i-1}},\Delta \Ec^{\phi}_{\tau

book defined in Section \[sec: general order book\ ] represent the moral force of the stock. Whenever the trader buys / sell $ n$ unit of stocks (being by an aggressive order or by the murder of a terminus ad quem order), she sells / buys immediately $ n$ units of the futures. Her inventory should be in full liquidated at $ T$. The optimal dominance trouble { # the - optimal - control - trouble } --------------------------- We assume that the citation price $ F$ of the futures is given by $ $ \begin{aligned } \label{eq: def F } F=\frac{P^{\mathfrak b}+P^{\mathfrak a}}{2}+S \end{aligned}$$ where $ S$ is a mean - revert process $ $ \begin{aligned } \label{eq: dyna S } S= S_{0}+\int_{0}^{\cdot}\rho(\hat s - S_t)dt+\int_{0}^{\cdot}\sigma(S_{t }) dW_t.\end{aligned}$$ Here, $ \rho$ is the strength of mean reversion, $ \hat s$ is the average of average reversion and $ \sigma:{\mathbb{R}}\mapsto { \mathbb{R}}$ is a Lipschitz bounded function representing the volatility of the procedure. In the following, we shall assume that $ $ \begin{aligned } \label{eq: ass sigma bounded support } \mbox { the support of $ \sigma$ is bounded.}\end{aligned}$$ \[rem: DS\ ] The above implies in particular that $ S$ lies in a certain compact set $ D_{S}$ as soon as $ S_{0}$ does. [ This could clearly be relaxed to the price of a finer psychoanalysis. ] { } The scheme of the HFT is described by the same quantities as the one of the MM in Section \[sec: marketmaker\ ]. The lone difference lies in the fact that she constantly holds a numeral equal to $ -I$ units of the futures $ F$. We assume that buying / sell the futures leads to the payment of a proportional cost $ \kappa\ge 0$. Then, the dynamics of the corresponding country process $ X$ is given by $ $ \begin{aligned } \label{eq: dyna X HFT } X_{\tau_{i}}=\Tc_{X}(S_{\tau_{i}},P_{\tau_{i-1}},Q_{\tau_{i-1}},X_{\tau_{i-1}},\Delta \Ec^{\phi}_{\tau

bolk defined in Section \[seg: general order yiok\] re'resenta the dyvamics of the stock. Whenever tye trqder buys/sells $n$ units of stockd (being vy ai aggressive ordxd or by the esccutimi of a limit orcer), she senls/buys immedidtdlv $n$ units of the futures. Her inventowy shoukd be fully liquydattd wt $T$. Fhe optimal control problem {#the-optjmal-conurol-problem} --------------------------- We assuke that the reference pricf $F$ lf the futures is hiven by $$\betin{ajugned} \label{eq: def F} F=\frab{'^{\mathfrak b}+K^{\mathfrak a}}{2}+S \end{aligned}$$ where $S$ ks a kean-revertunt pgmcess $$\begin{eligneq} \label{eq: dyna S} S= S_{0}+\ind_{0}^{\cdot}\rhp(\hat s-S_t)dt+\int_{0}^{\ccot}\vigna(S_{t}) dW_t.\end{aligned}$$ Hece, $\rho$ is the strengjh of mean rzversion, $\hat s$ is the aceragg of kean eevdrsjoi ahd $\sigla:{\methbb{R}}\mapstk {\mathbb{R}}$ iw a Lipschitz boundtd slmction repreaentind ehe volatility of the process. In the fonlosing, we shall assume thqt $$\begin{aligned} \label{ee: ass sigia bounded support} \mbox{ the support of $\sigma$ is bognded.}\xna{alnnked}$$ \[rdn : DS\] The above implies in particular that $S$ liqa on a certain comiact set $D_{S}$ as soom ws $F_{0}$ does. [This cuuld cleadly be relaxed to hhe prise of a finer wnalusis. ]{} The strategy of the HFR is describvd bt the same quantitnes as the oue of jhe MM in Section \[sec:marketmayer\]. Fhe only divference mkes in the fact ghan sha constantly holds a numbew equal ti $-I$ bnits of the futurqs $F$. We asdume bvat buying/selling hhe fotures leads to hhe payment of a proportional cost $\kappa\ge 0$. Tnet, tve dynamncs of the correspogding state prpcess $X$ is gkven by $$\befin{aligied} \label{eq: drna X HFT} X_{\tag_{l}}=\Tc_{X}(S_{\tau_{i}},P_{\tan_{i-1}},Q_{\tau_{i-1}},X_{\twu_{i-1}},\Dwlta \Ec^{\phi}_{\txj

book defined in Section \[sec: general order the of the Whenever the trader (being an aggressive order by the execution a limit order), she sells/buys immediately units of the futures. Her inventory should be fully liquidated at $T$. The control problem {#the-optimal-control-problem} --------------------------- We assume that the reference price $F$ of the is by \label{eq: F} F=\frac{P^{\mathfrak b}+P^{\mathfrak a}}{2}+S \end{aligned}$$ where $S$ is a mean-reverting process $$\begin{aligned} \label{eq: dyna S} S= s-S_t)dt+\int_{0}^{\cdot}\sigma(S_{t}) dW_t.\end{aligned}$$ Here, $\rho$ is the strength of reversion, $\hat s$ is average of mean reversion and {\mathbb{R}}$ a Lipschitz function the of the process. the following, we shall assume that $$\begin{aligned} \label{eq: ass sigma bounded support} \mbox{ the support of $\sigma$ bounded.}\end{aligned}$$ \[rem The above in that lies in a set $D_{S}$ as soon as $S_{0}$ clearly be relaxed to the price of a analysis. ]{} strategy of the HFT is described the same quantities as the one of the in Section \[sec:marketmaker\]. The only difference lies in the fact that she constantly holds a to $-I$ units of futures $F$. We that the leads the payment a proportional cost $\kappa\ge 0$. Then, the dynamics of the corresponding process $X$ is given by $$\begin{aligned} \label{eq: dyna X HFT}

book defined in Section \[sec: geNeral order Book\] rEprEseNtS the DynaMics of the stock. wHeneVer the trader buys/sells $n$ Units Of STockS (BeIng by An aggreSSiVE OrdEr Or By tHe EXeCutioN of A limit oRder), she selLs/bUyS immediately $N$ UnIts of the fuTurEs. Her inventoRy sHould bE fUllY LiquiDatEd at $T$. the optIMal conTrol problEm {#THe-optiMAl-contrOL-PrObleM} --------------------------- We assume that the rEFeREnce price $F$ of thE futurEs IS gIVEn bY $$\beGin{aligned} \LaBel{eq: DEf F} F=\fraC{p^{\mATHFraK B}+P^{\mathfrak a}}{2}+S \eNd{aligned}$$ whERe $S$ Is a meaN-rEveRTing prOcess $$\BeGIn{aLigned} \label{Eq: dyNa S} S= S_{0}+\int_{0}^{\cDot}\rho(\HAt s-S_t)dt+\INt_{0}^{\cdot}\sIgma(S_{t}) DW_t.\End{AligNEd}$$ heRe, $\rHo$ IS thE StRenGTh oF mean revErSiOn, $\hat S$ is tHE AVEragE of Mean ReverSion and $\sigma:{\mAthBb{R}}\mAPstO {\mathBb{R}}$ is A LipScHitz bOunded FunctIoN representing thE volAtility of The PrOceSs. in the FOllowiNg, wE shAll assuMe that $$\bEGin{AlIGNEd} \Label{eq: ass sigma bouNdED SuPport} \mboX{ the suPPoRt OF $\sigma$ is BoUndEd.}\enD{ALigneD}$$ \[rem : ds\] THe above iMplies IN pArTicular ThAt $S$ lieS iN a cErtAin coMPact Set $D_{S}$ aS soon as $S_{0}$ Does. [THIs could clearly BE relaxed to the PRiCE Of A FineR anAlysis. ]{} The stRateGY of tHe HFt Is DesCRibed By the SaME qUAntities as the one of tHe mM in SeCtion \[Sec:marketmakeR\]. The only diFFERence lieS in tHE fACt that she constAntly Holds a numbER equal to $-i$ unitS of the fuTures $F$. We aSSUme that bUyiNg/sEllIng THE fUtures leads to THE payMeNt of a prOpoRtional CosT $\kaPpa\Ge 0$. THeN, the dynamIcs of the CoRrEsPoNdiNg staTE process $x$ iS giVeN by $$\Begin{ALigned} \Label{Eq: dyNa x Hft} X_{\tAu_{i}}=\Tc_{X}(S_{\TAu_{I}},p_{\Tau_{i-1}},q_{\tAu_{I-1}},X_{\taU_{i-1}},\DElTa \Ec^{\pHi}_{\taU

book defined in Section \ [sec: gene ral o rde r b oo k\]repr esents the dyn a mics of the stock. Wheneve r the t r ader bu ys/se lls $n$ un i t s o fst ock s( be ing b y a n aggre ssive orde r o rby the execu t io n of a lim itorder), shesel ls/buy simm e diate ly$n$ u nits o f the f utures. H er invent o ry shou l d b e fu lly liquidated at $T $ . The optimal contr ol pr o b lem {# the-optima l- contr o l-probl e m} - - --- - ------------- -------- W e as sume t ha t t h e refe rence p r ice $F$ of the fut ures is g iven b y $$\beg i n{align ed} \l abe l{e q: d e fF} F= \f r ac{ P ^{ \ma t hfr ak b}+P^ {\ ma thfra k a} } { 2 } +S \ end {ali gned} $$ where $S$isa me a n-r evert ing p roce ss $$\b egin{a ligne d} \label{eq: dyn a S} S= S_{0} +\i nt _{0 }^ {\cdo t }\rho( \ha t s -S_t)dt +\int_{ 0 }^{ \c d o t }\ sigma(S_{t}) dW_t. \e n d {a ligned}$ $ Here , $ \r h o$ is th estr engt h of me an r e ve rsion, $ \hat s $ i sthe ave ra ge ofme anrev ersio n and $\sig ma:{\mat hbb{R } }\mapsto {\mat h bb{R}}$ is aL ip s c hi t z bo und ed function rep r esen ting th e v o latil ity o ft he process. In the fol lo wing,we sh all assume th at $$\begi n { a ligned}\lab e l{ e q: ass sigma b ounde d support} \mbox{ t he su pport of $\sigma$ i s bounde d.} \en d{a lig n e d} $$ \[rem : D S \ ] Th eabove i mpl ies inpar tic ula r t ha t $S$ lie s in a c er ta in c omp act s e t $D_{S} $asso onas $S _ {0}$ d oes.[Thi sco u ldclearly be r elax ed t o th e p ri ce of a f i ner analys is. ]{} The stra te gy of the HFT is descr ib ed by thesa mequanti t i es as th e one of the MM in Sect i on \[se c:m arket make r\]. Theonl y diff ere n ce lie s in t he fa ct th a t shec o ns tan tl y holds an u mbe r equ al to$-I$ un its of the futures $F$ . We assume t hat buy i n g/ sel l in g th ef utu r e s leads to thepayment of a pr oportional cos t$\kappa \ge 0$. Then , the dy namics of the corr es pond i n g s tate proce ss $X$ i s given b y $$\b e gi n{ali gne d} \la be l{e q: dy na X H F T} X_{\ tau_{i }} =\Tc_{ X}(S_ {\ tau_{i}} ,P_{\tau_{i-1}},Q_{\tau _{i-1} },X_{ \ta u_{i-1}}, \De l ta\Ec^{\phi }_{\ tau

book_defined in_Section \[sec: general order_book\] represents_the_dynamics of_the_stock. Whenever the_trader buys/sells $n$_units of stocks (being_by an aggressive_order_or by the execution of a limit order), she sells/buys immediately $n$ units of_the_futures. Her_inventory_should_be fully liquidated at $T$. The_optimal control problem {#the-optimal-control-problem} --------------------------- We assume_that the_reference price $F$ of the futures is given_by_$$\begin{aligned} \label{eq: def F} F=\frac{P^{\mathfrak_b}+P^{\mathfrak a}}{2}+S \end{aligned}$$ where $S$ is a mean-reverting process_$$\begin{aligned} \label{eq: dyna S} S= S_{0}+\int_{0}^{\cdot}\rho(\hat s-S_t)dt+\int_{0}^{\cdot}\sigma(S_{t}) dW_t.\end{aligned}$$_Here, $\rho$ is_the_strength_of mean reversion, $\hat_s$ is the average of mean_reversion and $\sigma:{\mathbb{R}}\mapsto {\mathbb{R}}$ is a_Lipschitz bounded function representing the volatility of_the process. In the following, we_shall assume that $$\begin{aligned} \label{eq: ass_sigma bounded_support} \mbox{ the support of $\sigma$_is bounded.}\end{aligned}$$ \[rem :_DS\] The_above implies in_particular that $S$ lies in a_certain compact set_$D_{S}$ as soon as $S_{0}$ does._[This_could clearly be_relaxed_to_the price_of a finer_analysis._]{} The strategy_of_the HFT is described by the_same_quantities as the one of the MM_in Section \[sec:marketmaker\]. The_only_difference lies in the_fact that she constantly holds_a number equal to $-I$ units_of the_futures $F$._We assume that buying/selling the futures leads to the payment of_a proportional cost $\kappa\ge 0$. Then,_the dynamics of the_corresponding state_process_$X$ is given_by_$$\begin{aligned} \label{eq: dyna_X HFT} X_{\tau_{i}}=\Tc_{X}(S_{\tau_{i}},P_{\tau_{i-1}},Q_{\tau_{i-1}},X_{\tau_{i-1}},\Delta \Ec^{\phi}_{\tau

24.94+0.07). The usable bandwidth of the 6.7 GHz maser observations is also [$\sim$]{}1.5 $\times$ greater (in [km s$^{-1}$]{}) than that of the 44 GHz maser observations (§\[vlaobs\],Tables \[67obstable\]-\[44obstable\]), and there are sources in which 44 GHz maser emission extends to, and most likely beyond, the edge of the observed velocity range (e.g. G10.29$-$0.13, Figure \[jcmtlineplots\]a, G18.89$-$0.47, Figure \[jcmtlineplots\]e). One generalization that may be made based on Figure \[jcmtlineplots\] is that 44 GHz Class I [CH$_3$OH]{} maser emission is invariably (but not exclusively) present at the gas [$v_{LSR}$]{} (as indicated by the [H$^{13}$CO$^{+}$]{} emission). The velocity of Class II 6.7 GHz [CH$_3$OH]{} maser emission, in contrast, exhibits every possible permutation with respect to the velocity of the thermal gas: coincident, anticorrelated, redshifted, blueshifted, and every combination of these characteristics. The relative velocities of the thermal gas and 6.7 GHz [CH$_3$OH]{} maser emission for each source are listed in Table \[summarytable\]. To investigate the spatial structure of the kinematic complexity of the maser spectra, Figure \[kinplots\] presents plots of maser positions color-coded by velocity. For each source, the left-hand panel of Fig. \[kinplots\] shows 44 GHz [CH$_3$OH]{} maser positions overplotted on the 4.5 [$\mu$m]{} image from the GLIMPSE survey. On the scale of these plots the uncertainty in the 44 GHz maser positions is smaller than the symbol size, and every fitted position from Table \[maserfitparams\_44\] is plotted individually. For each source, the black rectangle(s) overplotted on the 4.5 [$\mu$m]{} image is the field of view of the right-hand panel(s), which show the positions of 6.7 GHz masers color-coded by velocity. In the 6.7 GHz maser plots, the sizes of the crosses correspond to the relative positional uncertainties for each maser in Table \[maserfitparams

24.94 + 0.07). The usable bandwidth of the 6.7 GHz maser observations is also [ $ \sim$]{}1.5 $ \times$ great (in [ km   s$^{-1}$ ] { }) than that of the 44 gigahertz maser observations (§ \[vlaobs\],Tables   \[67obstable\]-\[44obstable\ ]), and there are source in which 44 gigahertz maser emission extends to, and most probable beyond, the boundary of the observed velocity range (e.g. G10.29$-$0.13, Figure   \[jcmtlineplots\]a, G18.89$-$0.47, Figure   \[jcmtlineplots\]e). One generalization that may be take based on Figure   \[jcmtlineplots\ ] is that 44 GHz course I [ CH$_3$OH ] { } maser emission is invariably (but not exclusively) present at the gas [ $ v_{LSR}$ ] { } (as indicated by the [ H$^{13}$CO$^{+}$ ] { } emission). The velocity of Class II 6.7 gigahertz [ CH$_3$OH ] { } maser emission, in contrast, exhibits every potential permutation with respect to the velocity of the thermal gas: coincident, anticorrelated, redshifted, blueshifted, and every combination of these characteristics. The proportional velocities of the thermal natural gas and 6.7 GHz [ CH$_3$OH ] { } maser emission for each source are list in Table   \[summarytable\ ]. To investigate the spatial structure of the kinematic complexity of the maser spectra, Figure   \[kinplots\ ] presents plots of maser positions color - gull by velocity. For each source, the left - hand panel of Fig.   \[kinplots\ ] shows 44 GHz [ CH$_3$OH ] { } maser positions overplotted on the 4.5 [ $ \mu$m ] { } image from the GLIMPSE survey. On the plate of these plot the uncertainty in the 44 gigahertz maser situation is small than the symbol size, and every fitted position from Table   \[maserfitparams\_44\ ] is plat individually. For each source, the black rectangle(s) overplotted on the 4.5 [ $ \mu$m ] { } image is the field of view of the veracious - hand panel(s), which show the positions of 6.7 GHz masers coloring material - coded by velocity. In the 6.7 GHz maser plots, the size of the cross correspond to the relative positional uncertainties for each maser in Table   \[maserfitparams

24.94+0.07). Thf usable bandwidth of tht 6.7 GHz maser obsetvqtions is alao [$\sim$]{}1.5 $\tkmes$ greater (in [km s$^{-1}$]{}) than thav of the 44 GHz maser observationr (§\[vlaobs\],Twbles \[67obsrablt\]-\[44obstable\]), and thecs are sources ln whnci 44 GHz maser emlssion extetds to, and mosd uinely beyond, the edge of the observeq velocotj range (e.g. G10.29$-$0.13, Fygurt \[jcitlihvpoots\]a, G18.89$-$0.47, Figure \[jcmtlineplots\]e). Ohe genegalization that msy be made based on Figure \[ucmtpineplots\] is that 44 GHz Class U [CH$_3$JY]{} maser emisrion is inncriably (but not exclusively) present at the gas [$r_{LSR}$]{} (as inducqtef by the [H$^{13}$CO$^{+}$]{} emisfion). The velocity of Class OI 6.7 GHz [CH$_3$OH]{} mssec emussion, in contrast, exiibits every possiblg permutathou with respect to the vwlocijy of the rhefmam jas: coincldeit, anticorrslated, redsyifted, blueshifted, snq every combinztion jf these characteristics. The relative velmcifies of the thermal gas and 6.7 GHz [CH$_3$OH]{} maser gmission fjr each source are listed in Table \[summarytable\]. To hnvesvieatt bme soqtlal structure of the kinematic complexity of ege msser spectra, Flgure \[kinplots\] presrnhs klots of maser positnkna color-coded by vepocity. Sor eqch sourct, the left-hand panel of Fig. \[kinplits\] shows 44 GKz [XH$_3$OH]{} maser positious overplottzd on jhe 4.5 [$\mi$m]{} image from the GLIMPRE shrvey. On thf scale or these plots the unbertdinty in the 44 GHz maser pofitions iw smcller thxn tne symfol size, ajd evcsy fitted position from Tdble \[maserfltparams\_44\] is plotted individuallb. For each soorca, tve black rectsngle(s) overpljtted on the 4.5 [$\ku$m]{} imaye is ghe field kf view of the rigrt-hand panel(s), ahich show tie positijns if 6.7 THz masdfs color-coded ny velociny. In rhe 6.7 GHz maser plobs, thg aizes of the crissts correspond to thd rqlwtmve pjvitional uncartakntkrs fof each masev iv Tanle \[maserfitparams

24.94+0.07). The usable bandwidth of the 6.7 observations also [$\sim$]{}1.5 greater (in [km 44 maser observations (§\[vlaobs\],Tables and there are in which 44 GHz maser emission to, and most likely beyond, the edge of the observed velocity range (e.g. Figure \[jcmtlineplots\]a, G18.89$-$0.47, Figure \[jcmtlineplots\]e). One generalization that may be made based on \[jcmtlineplots\] that GHz I [CH$_3$OH]{} maser emission is invariably (but not exclusively) present at the gas [$v_{LSR}$]{} (as indicated the [H$^{13}$CO$^{+}$]{} emission). The velocity of Class II GHz [CH$_3$OH]{} maser emission, contrast, exhibits every possible permutation respect the velocity the gas: anticorrelated, redshifted, blueshifted, every combination of these characteristics. The relative velocities of the thermal gas and 6.7 GHz [CH$_3$OH]{} maser for each listed in \[summarytable\]. investigate spatial structure of complexity of the maser spectra, Figure of maser positions color-coded by velocity. For each the left-hand of Fig. \[kinplots\] shows 44 GHz maser positions overplotted on the 4.5 [$\mu$m]{} image the GLIMPSE survey. On the scale of these plots the uncertainty in the 44 GHz is smaller than the size, and every position Table is individually. For source, the black rectangle(s) overplotted on the 4.5 [$\mu$m]{} image is field of view of the right-hand panel(s), which show the 6.7 masers color-coded by In the 6.7 GHz plots, sizes of the crosses the positional maser Table

24.94+0.07). The usable bandwidth of the 6.7 GHZ maser obseRvatiOns Is aLsO [$\sim$]{}1.5 $\TimeS$ greater (in [km s$^{-1}$]{}) tHAn thAt of the 44 GHz maser observaTions (§\[VlAObs\],TABlEs \[67obsTable\]-\[44obSTaBLE\]), anD tHeRe aRe SOuRces iN whIch 44 GHz mAser emissiOn eXtEnds to, and mosT LiKely beyond, The Edge of the obsErvEd veloCiTy rANge (e.g. g10.29$-$0.13, FiGure \[jCmtlinEPlots\]a, g18.89$-$0.47, Figure \[jcMtLIneploTS\]e). One geNERaLizaTion that may be made BAsED on Figure \[jcmtlIneploTs\] IS tHAT 44 GHZ ClAss I [CH$_3$OH]{} maSeR emisSIon is inVArIABLy (bUT not exclusiveLy) present at THe gAs [$v_{LSR}$]{} (As IndICated bY the [H$^{13}$cO$^{+}$]{} EMisSion). The veloCity Of Class II 6.7 gHz [CH$_3$Oh]{} Maser emISsion, in ContraSt, eXhiBits EVeRy PosSiBLe pERmUtaTIon With respEcT tO the vElocITY OF the TheRmal Gas: coIncident, anticOrrElatED, reDshifTed, blUeshIfTed, anD every CombiNaTion of these charActeRistics. ThE reLaTivE vElociTIes of tHe tHerMal gas aNd 6.7 GHz [CH$_3$oh]{} maSeR EMIsSion for each source aRe LIStEd in TablE \[summaRYtAbLE\]. To invesTiGatE the SPAtial StruCTuRe of the kInematIC cOmPlexity Of The masEr SpeCtrA, FiguRE \[kinPlots\] pResents pLots oF Maser positions COlor-coded by veLOcITY. FOR eacH soUrce, the left-Hand PAnel Of FiG. \[KiNplOTs\] shoWs 44 GHz [cH$_3$oh]{} mASer positions overploTtEd on thE 4.5 [$\mu$m]{} iMage from the GLiMPSE surveY. oN The scale Of thESe PLots the uncertaInty iN the 44 GHz masER positioNs is sMaller thAn the symbOL Size, and eVerY fiTteD poSITiOn from Table \[maSERfitPaRams\_44\] is pLotTed indiVidUalLy. FOr eAcH source, thE black reCtAnGlE(s) OveRplotTEd on the 4.5 [$\mU$m]{} ImaGe Is tHe fieLD of vieW of thE rigHt-HaND paNel(s), whiCH sHOW the PoSiTionS of 6.7 gHZ maseRs coLOr-cOded by vElocity. In The 6.7 ghz maSeR pLots, the Sizes of the croSsEs correspoNd To tHe relaTIVe positiOnal uncertainties for eacH Maser in tabLe \[masErfiTparams

24.94+0.07). The usable ba ndwidth of the6.7 GH zmase r ob servations isa lso[$\sim$]{}1.5 $\times$ grea te r (in [k m s$^ {-1}$]{ } )t h anth at of t h e44 GH z m aser ob servations (§ \[ vlaobs\],Tab l es  \[67obsta ble \]-\[44obsta ble \]), a nd th e re ar e s ource s in w h ich 44 GHz mase re missio n extend s to , an d most likely bey o nd , the edge of t he obs er v ed v elo cit y range (e .g . G10 . 29$-$0. 1 3, F i gur e  \[jcmtlinepl ots\]a, G18 . 89$ -$0.47 ,Fig u re \[j cmtli ne p lot s\]e). One gen eralizati on tha t may be made ba sed on Fi gur e \[ j cm tl ine pl o ts\ ] i s t h at44 GHz C la ss I [C H$_3 $ O H ] {} m ase r em issio n is invariab ly(but not excl usive ly)pr esent at th e gas [ $v_{LSR}$]{} (a s in dicated b y t he [H $^ {13}$ C O$^{+} $]{ } e mission ). Thev elo ci t y of Class II 6.7 GHz[C H $ _3 $OH]{} m aser e m is si o n, in co nt ras t, e x h ibits eve r ypossible permu t at io n withre spectto th e v eloci t y of the t hermal g as: c o incident, anti c orrelated, re d sh i f te d , bl ues hifted, and eve r y co mbin a ti ono f the se ch ar a ct e ristics. The relati ve veloc ities of the therm al gas and 6 . 7 GHz [C H$_3 $ OH ] {} maser emiss ion f or each so u rce areliste d in Tab le \[summ a r ytable\] . Toinv est i g at e the spatial s truc tu re of t hekinemat iccom ple xit yof the ma ser spec tr a, F ig ure  \[ki n plots\]pr ese nt s p lotso f mase r pos itio ns c o lor -codedb yv e loci ty .Foreac hsourc e, t h e l eft-han d panel o f F i g. \ [k in plots\] shows 44 GHz [ CH$_3$OH]{ }mas er pos i t ions ove rplotted on the 4.5 [$\ m u$m]{}ima ge fr om t he GLIMPS E s urvey. On the sc ale of thes eplo t s theu n ce rta in ty in the4 4 GH z mas er pos itionsis smaller than th e sy mbol size, an d e very f it ted po s iti on fro m Table \[maserfi tparams\_4 4\ ] i s plottedi ndi vi dually. For ea ch so u rce, th e black r ectangle( s) ove r p lot ted on the 4.5 [$\ mu$m]{} i m age i s t he fi eld of vi ew of theright- h and pane l(s),wh ich sh ow th eposition s of 6.7 GHz masers col or-cod ed by ve locity. I n t h e 6 .7 GHz ma serplots, the si zes of t hec rosse s co r re spo n d tother elative p o si tio n a luncertainti e s for each ma s er inTabl e \[maserfitparam s

24.94+0.07). The_usable bandwidth_of the 6.7 GHz_maser observations_is_also [$\sim$]{}1.5_$\times$_greater (in [km s$^{-1}$]{})_than that of_the 44 GHz maser_observations (§\[vlaobs\],Tables \[67obstable\]-\[44obstable\]), and_there_are sources in which 44 GHz maser emission extends to, and most likely beyond,_the_edge of_the_observed_velocity range (e.g. G10.29$-$0.13, Figure \[jcmtlineplots\]a,_G18.89$-$0.47, Figure \[jcmtlineplots\]e). One generalization that may_be made_based on Figure \[jcmtlineplots\] is that 44 GHz Class_I_[CH$_3$OH]{} maser emission_is invariably (but not exclusively) present at the gas_[$v_{LSR}$]{} (as indicated by the [H$^{13}$CO$^{+}$]{}_emission). The velocity_of_Class_II 6.7 GHz [CH$_3$OH]{}_maser emission, in contrast, exhibits every_possible permutation with respect to the_velocity of the thermal gas: coincident, anticorrelated,_redshifted, blueshifted, and every combination of_these characteristics. The relative velocities_of the_thermal gas and 6.7 GHz_[CH$_3$OH]{} maser emission_for each_source are listed_in Table \[summarytable\]. To investigate the spatial structure_of the kinematic_complexity of the maser spectra, Figure \[kinplots\]_presents_plots of maser_positions_color-coded_by velocity._For each source,_the_left-hand panel_of_Fig. \[kinplots\] shows 44 GHz [CH$_3$OH]{} maser_positions_overplotted on the 4.5 [$\mu$m]{} image from_the GLIMPSE survey. On_the_scale of these plots_the uncertainty in the 44_GHz maser positions is smaller than_the symbol_size, and_every fitted position from Table \[maserfitparams\_44\] is plotted individually. For each source,_the black rectangle(s) overplotted on the_4.5 [$\mu$m]{} image is_the field_of_view of the_right-hand_panel(s), which_show the positions of 6.7 GHz masers_color-coded by_velocity. In the 6.7 GHz maser_plots, the sizes of_the_crosses correspond to the relative positional_uncertainties for each maser in Table \[maserfitparams

4}$ and \[0.48,1\] 2.4 $\times 10^{-3}$ ----------------------------------------- -------------------------- : Approximate Hausdorff error of the standard area-preserving method corresponding to the tests in Figure \[Jaklic\]. \[Tab2\] ![Two area-preserving $G^1$ polynomials interpolate (\[JaklicFunc\]) joining at $t=0.3678$ on the left panel and $t=0.48$ on the right panel.[]{data-label="Jaklic"}](Fig5.jpg "fig:"){width="60mm" height="60mm"}![Two area-preserving $G^1$ polynomials interpolate (\[JaklicFunc\]) joining at $t=0.3678$ on the left panel and $t=0.48$ on the right panel.[]{data-label="Jaklic"}](Fig6.jpg "fig:"){width="60mm" height="60mm"} Optimized Area-Preservation {#SecOptBez} --------------------------- In this next example we showcase the accuracy that may be gained by selecting optimized $r_1$ and $r_2$. We interpolate the function $f(x)=4x(x-0.5)(x-1)e^x$ on the interval $x\in[0,1]$. Figure \[OptimizedBezier\] shows $f(x)$, the standard cubic Hermite polynomial interpolant (see [@Bartel]), the standard area-preserving cubic Bézier with $P=0$ and then the optimized area-preserving Bézier. We note that Figure \[OptimizedBezier\] appears to only have 3 curves because the optimized curve is nearly identical to the target function $f(x)$. ![Interpolation of $f(x)=4x(x-0.5)(x-1)e^x$.[]{data-label="OptimizedBezier"}](Fig7.jpg){width="100mm" height="60mm"} Piecewise Interpolation I ------------------------- In the next example we construct a piecewise interpolation of the function $f(x)=(x+1)e^{x}-1$ on the interval $[0,1]$. We partition the domain into finer and finer subintervals, compute the interpolants on each subinterval, then compute the maximum $L^{\infty}$ error over

4}$ and \[0.48,1\ ] 2.4 $ \times 10^{-3}$ ----------------------------------------- -------------------------- : Approximate Hausdorff error of the standard area - preserving method correspond to the examination in Figure \[Jaklic\ ]. \[Tab2\ ] ! [ Two area - preserving $ G^1 $ polynomials interpolate (\[JaklicFunc\ ]) join at $ t=0.3678 $ on the left panel and $ t=0.48 $ on the correct panel.[]{data - label="Jaklic"}](Fig5.jpg " fig:"){width="60 mm " height="60mm"}![Two sphere - preserving $ G^1 $ polynomials interpolate (\[JaklicFunc\ ]) joining at $ t=0.3678 $ on the left panel and $ t=0.48 $ on the correct panel.[]{data - label="Jaklic"}](Fig6.jpg " fig:"){width="60 mm " height="60 mm " } Optimized Area - Preservation { # SecOptBez } --------------------------- In this next exemplar we showcase the accuracy that may be gain by selecting optimized $ r_1 $ and $ r_2$. We interpolate the function $ f(x)=4x(x-0.5)(x-1)e^x$ on the interval $ x\in[0,1]$. Figure \[OptimizedBezier\ ] shows $ f(x)$, the standard cubic Hermite polynomial interpolant (attend [ @Bartel ]), the standard area - preserving cubic Bézier with $ P=0 $ and then the optimized sphere - preserving Bézier. We note that trope \[OptimizedBezier\ ] appears to merely have 3 curves because the optimized curve is nearly identical to the prey function $ f(x)$. ! [ Interpolation of $ f(x)=4x(x-0.5)(x-1)e^x$.[]{data - label="OptimizedBezier"}](Fig7.jpg){width="100 mm " height="60 mm " } Piecewise Interpolation I ------------------------- In the next example we construct a piecewise interpolation of the function $ f(x)=(x+1)e^{x}-1 $ on the interval $ [ 0,1]$. We partition the domain into finer and finer subintervals, compute the interpolants on each subinterval, then compute the maximum $ L^{\infty}$ error over

4}$ and \[0.48,1\] 2.4 $\times 10^{-3}$ ----------------------------------------- -------------------------- : Approsimate Hxusdorff error of the standacd aeea-prtferving method corfespondinh to the tesus in Figure \[Jaklmd\]. \[Tab2\] ![Two area-lveserriig $G^1$ polynomialx interpoldte (\[JaklicFunc\]) juiuing at $t=0.3678$ on the left panel and $t=0.48$ on the ribhh panel.[]{data-labgl="Jakkyc"}](Fif5.jpg "fig:"){width="60mm" height="60mm"}![Two area-prsservinj $G^1$ polynomials interpolate (\[JaklicFunc\]) jolninh at $t=0.3678$ on the left panel and $j=0.48$ on rhe right pavel.[]{data-labtl="Laklic"}](Fig6.jpf "fig:"){width="60mm" height="60mm"} Optimized Xrea-Pxeservation {#SwcOojBez} --------------------------- In this iext eqample we showcase tha accurscy that may bc gaiied vy selecting optimizev $r_1$ and $r_2$. We interpojate the xuuction $f(x)=4x(x-0.5)(x-1)e^x$ on the ibtervdl $x\hn[0,1]$. Fkturd \[OktikiaedBezler\] shows $f(x)$, fhe standare cubic Hermite polunjnial interpolznt (seq [@Fartel]), the standard area-preserving cubib Bézjer with $P=0$ and then the optimized area-preserging Béziew. We note that Figure \[OptimizedBezier\] appears to mnly iaxe 3 cmrver bfcause the optimized curve is nearly identicaj tp nhe target functijn $f(x)$. ![Interpplwtojn of $f(x)=4x(x-0.5)(x-1)e^x$.[]{aata-laysl="KptimizedBezier"}](Fig7.upg){widtr="100mm" hwight="60mm"} Pitcewixe Interpolation I ------------------------- In the nezt example wv cobstruct a piecewisz interpolatnon of the gunction $f(x)=(x+1)e^{x}-1$ on the iutervam $[0,1]$. We partihion the sumain into finer anc xiner subintervals, compute the intecpolauts on exch xubintqrval, then compmde the maximum $L^{\invty}$ etror oeer

4}$ and \[0.48,1\] 2.4 $\times 10^{-3}$ ----------------------------------------- Approximate error of standard area-preserving method Figure \[Tab2\] ![Two area-preserving polynomials interpolate (\[JaklicFunc\]) at $t=0.3678$ on the left panel $t=0.48$ on the right panel.[]{data-label="Jaklic"}](Fig5.jpg "fig:"){width="60mm" height="60mm"}![Two area-preserving $G^1$ polynomials interpolate (\[JaklicFunc\]) joining $t=0.3678$ on the left panel and $t=0.48$ on the right panel.[]{data-label="Jaklic"}](Fig6.jpg "fig:"){width="60mm" height="60mm"} Area-Preservation --------------------------- this example we showcase the accuracy that may be gained by selecting optimized $r_1$ and $r_2$. We the function $f(x)=4x(x-0.5)(x-1)e^x$ on the interval $x\in[0,1]$. Figure shows $f(x)$, the standard Hermite polynomial interpolant (see [@Bartel]), standard cubic Bézier $P=0$ then optimized area-preserving Bézier. note that Figure \[OptimizedBezier\] appears to only have 3 curves because the optimized curve is nearly identical the target ![Interpolation of height="60mm"} Interpolation ------------------------- In the we construct a piecewise interpolation of on the interval $[0,1]$. We partition the domain finer and subintervals, compute the interpolants on each then compute the maximum $L^{\infty}$ error over

4}$ and \[0.48,1\] 2.4 $\times 10^{-3}$ ----------------------------------------- -------------------------- : Approximate HausdoRff error of The stAndArd ArEa-prEserVing method corrESponDing to the tests in Figure \[jakliC\]. \[TAB2\] ![Two AReA-presErving $G^1$ POlYNOmiAlS iNteRpOLaTe (\[JakLicfunc\]) joiNing at $t=0.3678$ on tHe lEfT panel and $t=0.48$ on THe Right panel.[]{DatA-label="Jaklic"}](fig5.Jpg "fig:"){WiDth="60MM" heigHt="60mM"}![Two aRea-preSErving $g^1$ polynomiAlS InterpOLate (\[JakLICFUnc\]) jOining at $t=0.3678$ on the lefT PaNEl and $t=0.48$ on the rigHt paneL.[]{dATa-LABel="jakLic"}](Fig6.jpg "fIg:"){Width="60MM" height="60MM"} OPTIMizED Area-PreservaTion {#SecOptBEZ} --------------------------- In This neXt ExaMPle we sHowcaSe THe aCcuracy that May bE gained by SelectINg optimIZed $r_1$ and $R_2$. We intErpOlaTe thE FuNcTioN $f(X)=4X(x-0.5)(x-1)E^X$ oN thE IntErval $x\in[0,1]$. fiGuRe \[OptImizEDbEZier\] ShoWs $f(x)$, The stAndard cubic HeRmiTe poLYnoMial iNterpOlanT (sEe [@BarTel]), the StandArD area-preserving CubiC Bézier wiTh $P=0$ AnD thEn The opTImized AreA-prEservinG Bézier. wE noTe THAT FIgure \[OptimizedBeziEr\] APPeArs to onlY have 3 cURvEs BEcause thE oPtiMizeD CUrve iS neaRLy IdenticaL to the TArGeT functiOn $F(x)$. ![InteRpOlaTioN of $f(x)=4X(X-0.5)(x-1)e^x$.[]{Data-laBel="OptimIzedBEZier"}](Fig7.jpg){widtH="100Mm" height="60mm"} PieCEwISE INTerpOlaTion I ------------------------- In the nExt eXAmplE we cONsTruCT a pieCewisE iNTeRPolation of the functiOn $F(x)=(x+1)e^{x}-1$ oN the iNterval $[0,1]$. We partItion the doMAIN into finEr anD FiNEr subintervals, CompuTe the interPOlants on Each sUbintervAl, then comPUTe the maxImuM $L^{\iNftY}$ erROR oVer

4}$ and \[0 .48,1\] 2. 4 $\ times 10^{-3}$ -- ---------------------- ----- -- - ---- - -- -- -- ------- - -- - - --- -- -- --- -- : Ap pro ximateHausdorfferr or of the stan d ar d area-pre ser ving methodcor respon di ngt o the te sts i n Figu r e \[Ja klic\]. \ [T a b2\] ! [Two ar e a -p rese rving $G^1$ polyn o mi a ls interpolate (\[Ja kl i cF u n c\] ) j oining at$t =0.36 7 8$ on t h el e f t p a nel and $t=0. 48$ on ther igh t pane l. []{ d ata-la bel=" Ja k lic "}](Fig5.jp g "f ig:"){wid th="60 m m" heig h t="60mm "}![Tw o a rea -pre s er vi ng$G ^ 1$p ol yno m ial s interp ol at e (\[ Jakl i c F u nc\] ) j oini ng at $t=0.3678$ o n t he l e ftpanel and$t=0 .4 8$ on the r ightpa nel.[]{data-lab el=" Jaklic"}] (Fi g6 .jp g"fig: " ){widt h=" 60m m" heig ht="60m m "} O p t i mi zed Area-Preservat io n {# SecOptBe z} --- - -- -- - -------- -- --- ---- - In t hisn ex t exampl e we s h ow ca se theac curacy t hat ma y beg aine d by s electing opti m ized $r_1$ and $r_2$. We int e rp o l at e the fu nction $f(x )=4x ( x-0. 5)(x - 1) e^x $ on t he in te r va l $x\in[0,1]$. Figur e\[Opti mized Bezier\] show s $f(x)$,t h e standar d cu b ic Hermite polyno mialinterpolan t (see [@ Barte l]), the standard a rea-pres erv ing cu bic B éz ier with $P=0 $ andth en theopt imizedare a-p res erv in g Bézier. We note t ha tFi gur e \[O p timizedB ez ier \] ap pears to onl y hav e 3cu rv e s b ecauset he o ptim iz ed cur veis near ly i d ent ical to the targ etf unct io n$f(x)$. ![Interpola ti on of $f(x )= 4x( x-0.5) ( x -1)e^x$. []{data-label="Optimize d Bezier" }]( Fig7. jpg) {width="1 00m m" hei ght = "60mm" } Pie cewis eInt e r polat i o nI - -- ---------- - - --- ----- -- In the ne xt example we cons t ruc t a piecewise in terp o l at ion of the f u nct i o n $f(x)=(x+1)e^ {x}-1$ onth e i nterval $[ 0 ,1] $. We par titionthe d o main in to finerand finer s ubin t e rva ls, comput e the in terpolant s on e a ch subi nte rval,th encompu te the max imum$L^{\i nf ty}$ e rrorov er

4}$ _ _ _ __ __ and \[0.48,1\]_ _ _ __ 2.4 $\times 10^{-3}$ ----------------------------------------- -------------------------- :_Approximate_Hausdorff error_of_the_standard area-preserving method corresponding to_the tests in Figure \[Jaklic\]._\[Tab2\] ![Two area-preserving_$G^1$ polynomials interpolate (\[JaklicFunc\]) joining at $t=0.3678$ on_the_left panel and_$t=0.48$ on the right panel.[]{data-label="Jaklic"}](Fig5.jpg "fig:"){width="60mm" height="60mm"}![Two area-preserving $G^1$_polynomials interpolate (\[JaklicFunc\]) joining at $t=0.3678$_on the left_panel_and_$t=0.48$ on the right_panel.[]{data-label="Jaklic"}](Fig6.jpg "fig:"){width="60mm" height="60mm"} Optimized Area-Preservation {#SecOptBez} --------------------------- In this_next example we showcase the accuracy_that may be gained by selecting optimized_$r_1$ and $r_2$. We interpolate the_function $f(x)=4x(x-0.5)(x-1)e^x$ on the interval_$x\in[0,1]$. Figure_\[OptimizedBezier\] shows $f(x)$, the standard_cubic Hermite polynomial_interpolant (see_[@Bartel]), the standard_area-preserving cubic Bézier with $P=0$ and_then the optimized_area-preserving Bézier. We note that Figure_\[OptimizedBezier\]_appears to only_have_3_curves because_the optimized curve_is_nearly identical_to_the target function $f(x)$. ![Interpolation of $f(x)=4x(x-0.5)(x-1)e^x$.[]{data-label="OptimizedBezier"}](Fig7.jpg){width="100mm"_height="60mm"} Piecewise_Interpolation I ------------------------- In the next example we construct_a piecewise interpolation of_the_function $f(x)=(x+1)e^{x}-1$ on the_interval $[0,1]$. We partition the_domain into finer and finer subintervals,_compute the_interpolants on_each subinterval, then compute the maximum $L^{\infty}$ error over

in Figure \[fig:meta\_ex\] (left), for instance, the edge weight connecting $s_1$ and $s_2$ is the probability that a trajectory reaches $s_2$ before it can return to $s_1$, when executing $\pi_1$. These probabilities may be estimated either by sampling (Monte Carlo simulations), or computed analytically. The first approach is trivially implemented, with the usual advantages (e.g. parallelism, access to a black-box simulator is all is needed) and disadvantages (slow convergence); here we develop the latter, which leads to a concise set of linear problems to be solved, and sheds light on both the mathematical and computational structure. Since the bottlenecks partition the statespace into disjoint sets, the probabilities $\widetilde{P}(s,a,s')$ can be quickly computed in each [cluster]{}separately. Let $a$ be the action corresponding to executing a policy $\pi_{\ensuremath{\mathsf{c}}\xspace}$ in [cluster]{} ${\ensuremath{\mathsf{c}}\xspace}$. Then $$\widetilde{P}(s,a,s') = H_{s,s'},\quad \text{ for all } s,s'\in {\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace},$$ where $H$ is the [*minimal non-negative solution*]{}, for each $s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}$, to the linear system $$\label{eqn:trans_probs} H_{s,s'} = {P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s,s') + \sum_{s''\in{\accentset{\circ}{{\ensuremath{\mathsf{c}}\xspace}}}}{P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s,s'')H_{s'',s'},\quad s\in{\ensuremath{\mathsf{c}}\xspace}, s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}\,,$$ or in matrix-vector form, $$\label{eqn:trans_probs_matvec} \bigl(I - {P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(I-J)\bigr)H = {P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\

in Figure   \[fig: meta\_ex\ ] (left), for instance, the edge weight connect $ s_1 $ and $ s_2 $ is the probability that a trajectory pass $ s_2 $ before it can return to $ s_1 $, when carry through $ \pi_1$. These probability may be estimated either by sampling (Monte Carlo pretense), or calculate analytically. The first approach is trivially implemented, with the usual advantages (for example parallelism, access to a black - box simulator is all is needed) and disadvantage (slow convergence); here we originate the latter, which leads to a concise set of analogue problems to be solved, and sheds sparkle on both the mathematical and computational structure. Since the constriction partition the statespace into disjoint sets, the probability $ \widetilde{P}(s, a, s')$ can be quickly computed in each [ cluster]{}separately. Let $ a$ be the action corresponding to execute a policy $ \pi_{\ensuremath{\mathsf{c}}\xspace}$ in [ cluster ] { }   $ { \ensuremath{\mathsf{c}}\xspace}$. Then $ $ \widetilde{P}(s, a, s') = H_{s, s'},\quad \text { for all } s, s'\in { \ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace},$$ where $ H$ is the [ * minimal non - negative solution * ] { }, for each $ s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}$, to the linear system $ $ \label{eqn: trans_probs } H_{s, s' } = { P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s, s') + \sum_{s''\in{\accentset{\circ}{{\ensuremath{\mathsf{c}}\xspace}}}}{P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s, s'')H_{s'',s'},\quad s\in{\ensuremath{\mathsf{c}}\xspace }, s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}\,,$$ or in matrix - vector form, $ $ \label{eqn: trans_probs_matvec } \bigl(I - { P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(I - J)\bigr)H = { P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\

in Figure \[fig:meta\_ex\] (left), fov instance, the ebte weijht conhecting $r_1$ and $s_2$ is the probability tiat q trahectory reaches $s_2$ befofe it can return ro $s_1$, qhen execuvjng $\pi_1$. Bkese lvobabnlmties may be esjimated eithar by sampling (Munce Carlo simulations), or computed anajyticalky. The first apptoach ys tdpvlally implemented, with the usual advanteges (e.g. parallekism, access to a black-box dimupator is all is nefded) and diwadvwbtages (slow zonvergenct); kere we devglop the latter, which leads to a concnse set of oibeag problems ti be folved, and smvds ligvt on bpth the mathemstiral qnd computational strncture. Since the botjlenecks pdrcition the statespace ibto dhsjohnt rwts, tht pcobzbilitles $\widetilde{L}(s,a,s')$ can be quickly computed im qqch [cluster]{}selaratejy. Jet $a$ be the action corresponding to extcutihg a policy $\pi_{\ensurematy{\mathsf{c}}\xspace}$ in [cludter]{} ${\ensurqmath{\mathsf{c}}\xspace}$. Then $$\widetilde{P}(s,a,s') = H_{s,s'},\quad \taxt{ fkf aol } w,s'\ln {\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xs[zct},$$ wmere $H$ is the [*mikimal non-negative xopuyyon*]{}, for each $r'\in{\ensbdejath{\partial{\ensuremwth{\mathff{c}}\xspace}}\xspact}$, to yhe linear system $$\label{eqn:teans_probs} H_{s,s'} = {P_{\ensuremath{\mathsy{c}}\xspace}^{\pi_{\enrurekath{\msthsf{c}}\xspace}}}(s,s') + \sum_{s''\in{\aecentsst{\circ}{{\ensurfmath{\mathaw{c}}\xspace}}}}{P_{\ensuremxth{\kadhsf{c}}\xspace}^{\pi_{\ensuremath{\matrsf{c}}\xspacx}}}(s,s'')H_{s'',x'},\quad s\kn{\enxuremaeh{\mathsf{c}}\xdpace}, s'\in{\ensuremath{\partiwl{\ensorematv{\mathsf{c}}\xsoace}}\xspace}\,,$$ or in matrix-vector form, $$\label{eqn:trsnv_prmbs_matvee} \bigl(I - {P_{\ensuremath{\iathsf{c}}\xspace}^{\pn_{\ensuremcth{\matfsf{c}}\xspace}}}(P-J)\bigr)H = {'_{\ensuremath{\mwthsf{c}}\xspace}^{\ph_{\

in Figure \[fig:meta\_ex\] (left), for instance, the connecting and $s_2$ the probability that it return to $s_1$, executing $\pi_1$. These may be estimated either by sampling Carlo simulations), or computed analytically. The first approach is trivially implemented, with the advantages (e.g. parallelism, access to a black-box simulator is all is needed) and (slow here develop latter, which leads to a concise set of linear problems to be solved, and sheds light both the mathematical and computational structure. Since the partition the statespace into sets, the probabilities $\widetilde{P}(s,a,s')$ can quickly in each Let be action corresponding to a policy $\pi_{\ensuremath{\mathsf{c}}\xspace}$ in [cluster]{} ${\ensuremath{\mathsf{c}}\xspace}$. Then $$\widetilde{P}(s,a,s') = H_{s,s'},\quad \text{ for all } s,s'\in {\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace},$$ where is the solution*]{}, for $s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}$, the system $$\label{eqn:trans_probs} H_{s,s'} + \sum_{s''\in{\accentset{\circ}{{\ensuremath{\mathsf{c}}\xspace}}}}{P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s,s'')H_{s'',s'},\quad s\in{\ensuremath{\mathsf{c}}\xspace}, s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}\,,$$ or in \bigl(I - {P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(I-J)\bigr)H = {P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\

in Figure \[fig:meta\_ex\] (left), for iNstance, the Edge wEigHt cOnNectIng $s_1$ And $s_2$ is the probaBIlitY that a trajectory reacheS $s_2$ befOrE It caN ReTurn tO $s_1$, when eXEcUTIng $\Pi_1$. thEse PrOBaBilitIes May be esTimated eitHer By Sampling (MontE caRlo simulatIonS), or computed aNalYticalLy. the FIrst aPprOach iS triviALly impLemented, wItH The usuAL advantAGEs (E.g. paRallelism, access to A BlACk-box simulator Is all iS nEEdED) And DisAdvantages (SlOw conVErgence); HErE WE DevELop the latter, wHich leads to A ConCise seT oF liNEar proBlems To BE soLved, and shedS ligHt on both tHe mathEMatical ANd compuTationAl sTruCturE. siNcE thE bOTtlENeCks PArtItion the StAtEspacE intO DISJoinT seTs, thE probAbilities $\wideTilDe{P}(s,A,S')$ caN be quIckly CompUtEd in eAch [cluSter]{}sEpArately. Let $a$ be thE actIon corresPonDiNg tO eXecutINg a polIcy $\Pi_{\eNsuremaTh{\mathsF{C}}\xsPaCE}$ IN [cLuster]{} ${\ensuremath{\maThSF{C}}\xSpace}$. TheN $$\widetILdE{P}(S,A,s') = H_{s,s'},\quaD \tExt{ For aLL } S,s'\in {\eNsurEMaTh{\partiaL{\ensurEMaTh{\Mathsf{c}}\XsPace}}\xsPaCe},$$ wHerE $H$ is tHE [*minImal noN-negativE soluTIon*]{}, for each $s'\in{\eNSuremath{\partiAL{\eNSUrEMath{\MatHsf{c}}\xspace}}\xSpacE}$, To thE linEAr SysTEm $$\labEl{eqn:TrANs_PRobs} H_{s,s'} = {P_{\ensuremath{\mAtHsf{c}}\xsPace}^{\pI_{\ensuremath{\maThsf{c}}\xspacE}}}(S,S') + \Sum_{s''\in{\acCentSEt{\CIrc}{{\ensuremath{\mAthsf{C}}\xspace}}}}{P_{\enSUremath{\mAthsf{C}}\xspace}^{\pI_{\ensuremaTH{\Mathsf{c}}\xSpaCe}}}(s,S'')H_{s'',S'},\quAD S\iN{\ensuremath{\maTHSf{c}}\xSpAce}, s'\in{\eNsuRemath{\pArtIal{\EnsUreMaTh{\mathsf{c}}\Xspace}}\xsPaCe}\,,$$ Or In MatRix-veCTor form, $$\lAbEl{eQn:TraNs_proBS_matveC} \bigl(i - {P_{\enSuReMAth{\Mathsf{c}}\XSpACE}^{\pi_{\eNsUrEmatH{\maThSf{c}}\xsPace}}}(i-j)\biGr)H = {P_{\ensUremath{\maThsF{C}}\xspAcE}^{\pI_{\

in Figure \[fig:meta\_ex\ ] (left),for i nst anc e, the edg e weight conne c ting $s_1$ and $s_2$ is th e pro ba b ilit y t hat a trajec t or y rea ch es $s _2 $ b efore it can re turn to $s _1$ ,when executi n g$\pi_1$. T hes e probabilit ies may b eest i mated ei therby sam p ling ( Monte Car lo simula t ions),o r c ompu ted analytically. Th e first approac h is t ri v ia l l y i mpl emented, w it h the usual a d va n t a ges (e.g. paralle lism, acces s to a bla ck -bo x simul atoris all is needed) and disadvan tages( slow co n vergenc e); he rewedeve l op t hela t ter , w hic h le ads to a c on ciseseto f l inea r p robl ems t o be solved,and she d s l ighton bo th t he math ematic al an dcomputational s truc ture. Sin ceth e b ot tlene c ks par tit ion the st atespac e in to d i sj oint sets, the pro ba b i li ties $\w idetil d e{ P} ( s,a,s')$ c anbe q u i cklycomp u te d in eac h [clu s te r] {}separ at ely. Le t $ a$be th e act ion co rrespond ing t o executing a p o licy $\pi_{\e n su r e ma t h{\m ath sf{c}}\xspa ce}$ in [ clus t er ]{} ${\en surem at h {\ m athsf{c}}\xspace}$. T hen $$ \wide tilde{P}(s,a, s') = H_{s , s ' },\quad \te x t{ for all } s,s' \in { \ensuremat h {\partia l{\en suremath {\mathsf{ c } }\xspace }}\ xsp ace },$ $ wh ere $H$ is th e [*mi ni mal non -ne gativesol uti on* ]{} ,for each$s'\in{\ en su re ma th{ \part i al{\ensu re mat h{ \ma thsf{ c }}\xsp ace}} \xsp ac e} $ , t o the l i ne a r sys te m$$\l abe l{ eqn:t rans _ pro bs} H_{ s,s'} = {P_ { \ens ur em ath{\ma thsf{c}}\xspa ce }^{\pi_{\e ns ure math{\ m a thsf{c}} \xspace}}}(s,s') + \sum _ {s''\in {\a ccent set{ \circ}{{\ ens uremat h{\ m athsf{ c}}\xs pace} }} }{P _ { \ensu r e ma th{ \m athsf{c}}\ x s pac e}^{\ pi _{\e nsurema th{\mathsf{c}}\xsp a ce} }}(s,s'')H_{s '', s'}, \ q ua d s \ in { \en su r ema t h {\mathsf{c}}\xs pace}, s'\ in { \e nsuremath{ \ par ti al{\ens uremath {\mat h sf{c}}\ xspace}}\ xspace}\, ,$ $ or i n m atrix-vect or form, $$\label { eqn:t r an s_pro bs_ matvec }\bi gl(I- {P_{ \ ens urema th{\ma th sf{c}} \xspa ce }^{\pi_{ \ensuremath{\mathsf{c}} \xspac e}}}( I-J )\bigr)H= { P _{\ ensuremat h{\m athsf{c}}\ xsp ace }^{\p i_{ \

in_Figure \[fig:meta\_ex\] (left),_for instance, the edge_weight connecting_$s_1$_and $s_2$_is_the probability that_a trajectory reaches_$s_2$ before it can_return to $s_1$,_when_executing $\pi_1$. These probabilities may be estimated either by sampling (Monte Carlo simulations), or_computed_analytically. The_first_approach_is trivially implemented, with the_usual advantages (e.g. parallelism, access_to a_black-box simulator is all is needed) and disadvantages_(slow_convergence); here we_develop the latter, which leads to a concise set_of linear problems to be solved,_and sheds light_on_both_the mathematical and computational_structure. Since the bottlenecks partition the_statespace into disjoint sets, the probabilities_$\widetilde{P}(s,a,s')$ can be quickly computed in each_[cluster]{}separately. Let $a$ be the action corresponding_to executing a policy $\pi_{\ensuremath{\mathsf{c}}\xspace}$_in [cluster]{} ${\ensuremath{\mathsf{c}}\xspace}$._Then $$\widetilde{P}(s,a,s') = H_{s,s'},\quad _\text{ for all_} s,s'\in_{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace},$$ where $H$_is the [*minimal non-negative solution*]{}, for_each $s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}$, to_the linear system $$\label{eqn:trans_probs} H_{s,s'} = _{P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s,s')_+ \sum_{s''\in{\accentset{\circ}{{\ensuremath{\mathsf{c}}\xspace}}}}{P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(s,s'')H_{s'',s'},\quad s\in{\ensuremath{\mathsf{c}}\xspace},_s'\in{\ensuremath{\partial{\ensuremath{\mathsf{c}}\xspace}}\xspace}\,,$$_or_in matrix-vector_form, $$\label{eqn:trans_probs_matvec} \bigl(I -_{P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\ensuremath{\mathsf{c}}\xspace}}}(I-J)\bigr)H_= {P_{\ensuremath{\mathsf{c}}\xspace}^{\pi_{\

isation of the top-left $(n-|\Lambda|)\times (n-|\Lambda|)$ submatrix of $M_n$. In particular, after conditioning on any realisation of the top-left submatrix of this size, each of the $n-|\Lambda|$ equations above is satisfied with probability which is at most $\rho^{\F_p}(b)$, and the satisfaction of different equations is mutually independent. Hence, by the law of total probability, the system \[eqn:orthogonal-many-2\] is satisfied with probability at most $$\left(\rho^{\F_p}(b)\right)^{n-|\Lambda|} \leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-|\Lambda|}\leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-s_2},$$ where the middle bound follows from \[corollary:halasz-usable\], and the right-hand bound follows since $|\Lambda|\leq s_2$. Finally, substituting this in \[eqn:prob-orthog-many\] and proceeding as in \[eqn:easy-entropy-bound\] gives the desired conclusion. \[corollary:prob-level-set\] Let $p$ be a prime and $k\in \N, s_1,s_2,d\in [n], t\in [p]$ such that $1\\leq s_1 \leq s_2 \leq n/2$, $s_1 \leq d \leq n$, $p \leq \min\left\{e^{-s_1/2k}, \left(4k/s_1\right)^{k}\right\}$, and $t \geq s_1 \geq k \geq 100$. Then, for $0\leq \beta := \beta(n) \leq 1/2$, $$\Pr_{M_n}[\exists a \in \Bad^{d}_{k,s_1,s_2,t}(n)\setminus \Bad^{d}_{k,s_1,s_2,t+1}(n)|a \text{ is orthogonal to $\geq (1-\beta)n$ rows of } M_n] \leq (12C)^{n}p^{\beta n + 2s_2 + \frac{s_1

isation of the top - left $ (n-|\Lambda|)\times (n-|\Lambda|)$ submatrix of $ M_n$. In particular, after conditioning on any realisation of the top - left submatrix of this size, each of the $ n-|\Lambda|$ equation above is quenched with probability which is at most $ \rho^{\F_p}(b)$, and the satisfaction of unlike equations is mutually autonomous. Hence, by the law of full probability, the organization \[eqn: orthogonal - many-2\ ] is satisfied with probability at about $ $ \left(\rho^{\F_p}(b)\right)^{n-|\Lambda| } \leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-|\Lambda|}\leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n - s_2},$$ where the middle tie follows from \[corollary: halasz - usable\ ], and the correct - hand bound follow since $ |\Lambda|\leq s_2$. Finally, substituting this in \[eqn: prob - orthog - many\ ] and proceeding as in \[eqn: easy - entropy - bound\ ] gives the desire conclusion. \[corollary: prob - level - set\ ] Let $ p$ be a choice and $ k\in \N, s_1,s_2,d\in [ n ], t\in [ p]$ such that $ 1\\leq s_1 \leq s_2 \leq n/2 $, $ s_1 \leq d \leq n$, $ p \leq \min\left\{e^{-s_1/2k }, \left(4k / s_1\right)^{k}\right\}$, and $ t \geq s_1 \geq k \geq 100$. Then, for $ 0\leq \beta: = \beta(n) \leq 1/2 $, $ $ \Pr_{M_n}[\exists a \in \Bad^{d}_{k, s_1,s_2,t}(n)\setminus \Bad^{d}_{k, s_1,s_2,t+1}(n)|a \text { is orthogonal to $ \geq (1-\beta)n$ row of } M_n ] \leq (12C)^{n}p^{\beta n + 2s_2 + \frac{s_1

isahion of the top-left $(n-|\Lamnda|)\times (n-|\Lambda|)$ submatcix of $J_n$. In pafticular, after conditioning ln any eealisation of the top-ueft submwtrix of thiw size, each of the $k-|\Jambsw|$ eqbavions above is xatisfied fith probabilidy wkich is at most $\rho^{\F_p}(b)$, and the satissaction ov different eqoatiomf is mutually independent. Hence, by the law of total probabolity, the system \[eqn:orthoglnal-lany-2\] is satisfied aith probabulitr at most $$\lefg(\rho^{\F_p}(b)\right)^{n-|\Lambda|} \les \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-|\Laobda|}\lzq \left(\frac{2X(t+1)\wqrh{n}}{p\sqrt{s_1}}\righv)^{n-s_2},$$ whvre the middlc bound followx from \[corollavy:halesz-uwable\], and the right-haid bound follows sinse $|\Lambda|\nes s_2$. Finally, substutyting thiv in \[wqn:orog-octhkg-many\] anv proceedinf as in \[eqn:wasy-entropy-bound\] gifef the desired donclufijn. \[corollary:prob-level-set\] Let $p$ be a primt and $k\in \N, s_1,s_2,d\in [n], t\in [p]$ suxh that $1\\leq s_1 \leq s_2 \lgq n/2$, $s_1 \leq d \leq n$, $p \leq \min\left\{e^{-s_1/2k}, \left(4k/s_1\right)^{k}\right\}$, and $t \geq a_1 \gew h \tee 100$. Then, for $0\leq \beta := \beta(n) \leq 1/2$, $$\Pr_{M_n}[\exists a \jn \Bsd^{d}_{k,s_1,s_2,t}(n)\setminms \Bad^{d}_{k,s_1,s_2,t+1}(n)|a \text{ id ptthogonal to $\ggq (1-\beta)u$ rkws of } M_n] \leq (12C)^{n}p^{\heta n + 2s_2 + \feac{s_1

isation of the top-left $(n-|\Lambda|)\times (n-|\Lambda|)$ submatrix In after conditioning any realisation of size, of the $n-|\Lambda|$ above is satisfied probability which is at most $\rho^{\F_p}(b)$, the satisfaction of different equations is mutually independent. Hence, by the law of probability, the system \[eqn:orthogonal-many-2\] is satisfied with probability at most $$\left(\rho^{\F_p}(b)\right)^{n-|\Lambda|} \leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-|\Lambda|}\leq where middle follows \[corollary:halasz-usable\], and the right-hand bound follows since $|\Lambda|\leq s_2$. Finally, substituting this in \[eqn:prob-orthog-many\] and proceeding in \[eqn:easy-entropy-bound\] gives the desired conclusion. \[corollary:prob-level-set\] Let be a prime and \N, s_1,s_2,d\in [n], t\in [p]$ that s_1 \leq \leq $s_1 d \leq n$, \leq \min\left\{e^{-s_1/2k}, \left(4k/s_1\right)^{k}\right\}$, and $t \geq s_1 \geq k \geq 100$. Then, for $0\leq \beta := \beta(n) 1/2$, $$\Pr_{M_n}[\exists \Bad^{d}_{k,s_1,s_2,t}(n)\setminus \Bad^{d}_{k,s_1,s_2,t+1}(n)|a is to (1-\beta)n$ rows of \leq (12C)^{n}p^{\beta n + 2s_2 +

isation of the top-left $(n-|\LambdA|)\times (n-|\LamBda|)$ suBmaTriX oF $M_n$. IN parTicular, after coNDitiOning on any realisation oF the tOp-LEft sUBmAtrix Of this sIZe, EACh oF tHe $N-|\LaMbDA|$ eQuatiOns Above is Satisfied wIth PrObability whiCH iS at most $\rho^{\f_p}(b)$, And the satisfActIon of dIfFerENt equAtiOns is MutualLY indepEndent. HenCe, BY the laW Of total PRObAbilIty, the system \[eqn:orTHoGOnal-many-2\] is satiSfied wItH PrOBAbiLitY at most $$\lefT(\rHo^{\F_p}(b)\RIght)^{n-|\LaMBdA|} \LEQ \leFT(\frac{2C(t+1)\sqrt{k}}{p\Sqrt{s_1}}\right)^{n-|\lAmbDa|}\leq \lEfT(\frAC{2C(t+1)\sqrT{k}}{p\sqRt{S_1}}\RigHt)^{n-s_2},$$ where thE midDle bound fOllows FRom \[coroLLary:halAsz-usaBle\], And The rIGhT-hAnd BoUNd fOLlOws SIncE $|\Lambda|\lEq S_2$. FInallY, subSTITUtinG thIs in \[Eqn:prOb-orthog-many\] aNd pRoceEDinG as in \[Eqn:eaSy-enTrOpy-boUnd\] givEs the DeSired conclusion. \[CoroLlary:prob-LevEl-Set\] leT $p$ be a PRime anD $k\iN \N, s_1,S_2,d\in [n], t\iN [p]$ such tHAt $1\\lEq S_1 \LEQ s_2 \Leq n/2$, $s_1 \leq d \leq n$, $p \leq \mIn\LEFt\{E^{-s_1/2k}, \left(4k/S_1\right)^{K}\RiGhT\}$, And $t \geq s_1 \GeQ k \gEq 100$. ThEN, For $0\leQ \betA := \BeTa(n) \leq 1/2$, $$\Pr_{m_n}[\exisTS a \In \bad^{d}_{k,s_1,s_2,T}(n)\SetminUs \bad^{D}_{k,s_1,S_2,t+1}(n)|a \tEXt{ is OrthogOnal to $\geQ (1-\beta)N$ Rows of } M_n] \leq (12C)^{n}p^{\BEta n + 2s_2 + \frac{s_1

isation of the top-left $( n-|\Lambda |)\ti mes (n -| \Lam bda| )$ submatrix o f $M_ n$. In particular, aft er co nd i tion i ng on a ny real i sa t i onof t heto p -l eft s ubm atrix o f this siz e,ea ch of the $n - |\ Lambda|$ e qua tions aboveissatisf ie d w i th pr oba bilit y whic h is at most $\r ho ^ {\F_p} ( b)$, an d th e sa tisfaction of dif f er e nt equations i s mutu al l yi n dep end ent. Hence ,by th e law of to t a l pr o bability, the system \[e q n:o rthogo na l-m a ny-2\] is s at i sfi ed with pro babi lity at m ost $$ \ left(\r h o^{\F_p }(b)\r igh t)^ {n-| \ La mb da| }\ leq \l eft ( \fr ac{2C(t+ 1) \s qrt{k }}{p \ s q r t{s_ 1}} \rig ht)^{ n-|\Lambda|}\ leq \le f t(\ frac{ 2C(t+ 1)\s qr t{k}} {p\sqr t{s_1 }} \right)^{n-s_2} ,$$where the mi dd lebo und f o llowsfro m \ [coroll ary:hal a sz- us a b l e\ ], and the right-h an d bo und foll ows si n ce $ | \Lambda| \l eqs_2$ . Final ly,s ub stitutin g this in \ [eqn:pr ob -ortho g- man y\] andp roce edingas in \[ eqn:e a sy-entropy-bou n d\] gives the de s i re d con clu sion. \[co roll a ry:p rob- l ev el- s et\]Let $ p$ be a prime and $k\in \ N, s_1,s _2,d\ in [n], t\in[p]$ sucht h a t $1\\le q s_ 1 \ l eq s_2 \leq n/ 2$, $ s_1 \leq d \leq n$, $p \ leq \min \left\{e^ { - s_1/2k}, \l eft (4k /s_ 1 \ ri ght)^{k}\righ t \ }$,an d $t \g eqs_1 \ge q k \g eq100 $. Then, fo r $0\leq \ be ta : = \ beta( n ) \leq 1 /2 $,$$ \Pr _{M_n } [\exis ts a\in\B ad ^ {d} _{k,s_1 , s_ 2 , t}(n )\ se tmin us\B ad^{d }_{k , s_1 ,s_2,t+ 1}(n)|a \ tex t { is o rt hogonal to $\geq (1- \b eta)n$ row sof} M_n] \ leq (12C )^{n}p^{\beta n + 2s_2+ \frac{ s_1

isation of_the top-left_$(n-|\Lambda|)\times (n-|\Lambda|)$ submatrix of_$M_n$. In_particular,_after conditioning_on_any realisation of_the top-left submatrix_of this size, each_of the $n-|\Lambda|$_equations_above is satisfied with probability which is at most $\rho^{\F_p}(b)$, and the satisfaction of_different_equations is_mutually_independent._Hence, by the law of_total probability, the system \[eqn:orthogonal-many-2\]_is satisfied_with probability at most $$\left(\rho^{\F_p}(b)\right)^{n-|\Lambda|} \leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-|\Lambda|}\leq \left(\frac{2C(t+1)\sqrt{k}}{p\sqrt{s_1}}\right)^{n-s_2},$$_where_the middle bound_follows from \[corollary:halasz-usable\], and the right-hand bound follows since_$|\Lambda|\leq s_2$. Finally, substituting this in_\[eqn:prob-orthog-many\] and proceeding_as_in_\[eqn:easy-entropy-bound\] gives the desired_conclusion. \[corollary:prob-level-set\] Let $p$ be a prime_and $k\in \N, s_1,s_2,d\in [n], t\in_[p]$ such that $1\\leq s_1 \leq s_2_\leq n/2$, $s_1 \leq d \leq_n$, $p \leq \min\left\{e^{-s_1/2k}, \left(4k/s_1\right)^{k}\right\}$,_and $t_\geq s_1 \geq k \geq_100$. Then, for_$0\leq \beta_:= \beta(n) \leq_1/2$, $$\Pr_{M_n}[\exists a \in \Bad^{d}_{k,s_1,s_2,t}(n)\setminus \Bad^{d}_{k,s_1,s_2,t+1}(n)|a_\text{ is orthogonal_to $\geq (1-\beta)n$ rows of }_M_n]_\leq (12C)^{n}p^{\beta n_+_2s_2_+ \frac{s_1

$, which concludes the proof of the proposition. 0.3cm [**Proof of Propsition\[Prop Gamm 0 compact\]:**]{} The proof is similar as that of Proposition \[Prop weak convergence\] and easier. The proof will be omitted. Acknowledgment {#acknowledgment.unnumbered} ============== This work was supported by National Natural Science Foundation of China (NSFC) (No. 11431014, No. 11401557), and the Fundamental Research Funds for the Central Universities (No. 0010000048). [0]{} U. Baier and M.I. Freidlin, Theorems on large deviations and stability under random perturbations. [*Dokl. Akad. Nauk USSR*]{} [**235**]{}(1977) 253–256. A. Budhiraja and P. Dupuis, A variational representation for positive functionals of infinite dimensional Brownian motion. [*Probab. Math. Statist.*]{} [**20**]{}(2000) 39–61. A. Budhiraja, P. Dupuis and A. Ganguly, Moderate Deviation Principle for Stochastic Differential Equations with Jump, arXiv:1401.73v1. X. Chen, Probabilities of moderate deviations for B-valued independent random vectors. [*Chinese J. Contemporary Math.*]{} [**11**]{}(1990) 381–393. X. Chen, Probabilities of moderate deviations for independent random vectors in a Banach space. [*Chinese J. Appl. Probab. Statist.*]{} [**7**]{}(1991) 24–33. X. Chen, On lower bound of the moderate deviations of i.i.d. random variables in Banach space. [*Chinese J. Appl. Probab. Statist.*]{} [**9**]{}(1993) 386–393. I. Chueshov and A. Millet, Stochastic 2D Hydrodynamics Type Systems: Well Posedness and Large Deviations. [*Appl. Math. Optim.*]{} [**61**]{}(2010) 379–420. A. Dembo and O. Zeitouni, *Large deviations techniques and applications*, (2000) Springer-Verlag, New York. H. Djellout and A. Guillin, Moderate deviations for Markov chains with atom Stochastic processes and their applications. [*Stoch. Proc. Appl.*]{} [**95**]{}(2)(2001) 203–221. Z. Dong, J. X

$, which concludes the proof of the proposition. 0.3 cm [ * * validation of Propsition\[Prop Gamm 0 compact\ ] :* * ] { } The validation is alike as that of Proposition \[Prop weak convergence\ ] and easier. The validation will be omitted. Acknowledgment { # acknowledgment.unnumbered } = = = = = = = = = = = = = = This work was supported by National Natural Science Foundation of China (NSFC) (nobelium. 11431014, No. 11401557), and the Fundamental Research Funds for the Central Universities (nobelium. 0010000048). [ 0 ] { } U. Baier and M.I. Freidlin, Theorems on large deviations and stability under random perturbations. [ * Dokl. Akad. Nauk USSR * ] { } [ * * 235**]{}(1977) 253–256. A. Budhiraja and P. Dupuis, A variational theatrical performance for positive functionals of infinite dimensional Brownian motion. [ * Probab. Math. Statist. * ] { } [ * * 20**]{}(2000) 39–61. A. Budhiraja, P. Dupuis and A. Ganguly, Moderate Deviation Principle for Stochastic Differential Equations with Jump, arXiv:1401.73v1. X. Chen, Probabilities of moderate deviations for bel - valued autonomous random vectors. [ * Chinese J. Contemporary Math. * ] { } [ * * 11**]{}(1990) 381–393. X. Chen, Probabilities of moderate deviation for independent random vectors in a Banach space. [ * Chinese J. Appl. Probab. Statist. * ] { } [ * * 7**]{}(1991) 24–33. X. Chen, On lower adhere of the moderate deviations of i.i.d. random variables in Banach space. [ * Chinese J. Appl. Probab. Statist. * ] { } [ * * 9**]{}(1993) 386–393. I. Chueshov and A. Millet, Stochastic 2D Hydrodynamics Type Systems: Well Posedness and Large Deviations. [ * Appl. Math. Optim. * ] { } [ * * 61**]{}(2010) 379–420. A. Dembo and O. Zeitouni, * Large deviation techniques and applications *, (2000) Springer - Verlag, New York. H. Djellout and A. Guillin, Moderate deviations for Markov chains with atom Stochastic processes and their application. [ * Stoch. Proc. Appl. * ] { } [ * * 95**]{}(2)(2001) 203–221. Z. Dong, J. X

$, whlch concludes the proof uf the propositnin. 0.3cm [**Prmof of Propsitkon\[Prop Gamm 0 compact\]:**]{} The prlod is wimilar as that of Prooosition \[Irop weak contergence\] and easmsr. The ixoof slll bz imitted. Acknowlgdgment {#acknmwledgment.unnukbdrzd} ============== This work was supported by Nationaj Natursl Science Foundwtiom of Dhina (NSFC) (No. 11431014, No. 11401557), and the Fundamehtal Revearch Funds gor the Central Universitifs (Nl. 0010000048). [0]{} U. Baier and M.I. Fgeidlin, Theiremf on large dexiations and stability under random perturbations. [*Doku. Akab. Nauk USSR*]{} [**235**]{}(1977) 253–256. Q. Bkghiraja and P. Duiuis, A variatlpnal rapresenyation for posltive fubctionals of infinite dimensional Brownyan motiot. [*'robab. Math. Statist.*]{} [**20**]{}(2000) 39–61. Q. Vudhitaja, [. Duoyis ans E. Gznguly, Moverate Deviztion Princuple for Stochastic Dydferential Eqhationf rith Jump, arXiv:1401.73v1. X. Chen, Probabilities of moserate deviations for B-calued independent rajdom vectjrs. [*Chinese J. Contemporary Math.*]{} [**11**]{}(1990) 381–393. X. Chen, Probabilhties uf noqdeahe deviations for independent random vectors yh s Nanach space. [*Chikese J. Appl. Probab. Shayyst.*]{} [**7**]{}(1991) 24–33. X. Chen, Un lowzd gound of the moderwte devyatiobs of i.i.d. rancom variables in Banach spaxe. [*Chinese J. Qppl. Probab. Statisc.*]{} [**9**]{}(1993) 386–393. I. Chueshor and S. Milket, Stochastic 2D Hydrodvnamica Type Systfms: Well Lusedness and Laree Ceeiations. [*Appl. Math. Optim.*]{} [**61**]{}(2010) 379–420. W. Dembo aid O. Veitouni, *Larbe devyations tefhniqmas and applicationd*, (2000) Sptinger-Eerlag, New York. H. Djellout and A. Guillin, Moderate deviatoots xor Markjv chsins with atoi Stochastic ptocesses cnd thdir applicztions. [*Vtoch. Proc. Wppl.*]{} [**95**]{}(2)(2001) 203–221. Z. Dong, B. X

$, which concludes the proof of the [**Proof Propsition\[Prop Gamm compact\]:**]{} The proof Proposition weak convergence\] and The proof will omitted. Acknowledgment {#acknowledgment.unnumbered} ============== This work supported by National Natural Science Foundation of China (NSFC) (No. 11431014, No. 11401557), the Fundamental Research Funds for the Central Universities (No. 0010000048). [0]{} U. Baier M.I. Theorems large and stability under random perturbations. [*Dokl. Akad. Nauk USSR*]{} [**235**]{}(1977) 253–256. A. Budhiraja and P. Dupuis, variational representation for positive functionals of infinite dimensional motion. [*Probab. Math. Statist.*]{} 39–61. A. Budhiraja, P. Dupuis A. Moderate Deviation for Differential with Jump, arXiv:1401.73v1. Chen, Probabilities of moderate deviations for B-valued independent random vectors. [*Chinese J. Contemporary Math.*]{} [**11**]{}(1990) 381–393. X. Probabilities of for independent vectors a space. [*Chinese J. Statist.*]{} [**7**]{}(1991) 24–33. X. Chen, On the moderate deviations of i.i.d. random variables in space. [*Chinese Appl. Probab. Statist.*]{} [**9**]{}(1993) 386–393. I. and A. Millet, Stochastic 2D Hydrodynamics Type Systems: Posedness and Large Deviations. [*Appl. Math. Optim.*]{} [**61**]{}(2010) 379–420. A. Dembo and O. Zeitouni, *Large and applications*, (2000) Springer-Verlag, York. H. Djellout A. Moderate for chains with Stochastic processes and their applications. [*Stoch. Proc. Appl.*]{} [**95**]{}(2)(2001) 203–221. Z. J. X

$, which concludes the proof of tHe propositIon. 0.3cm [**proOf oF PRopsItioN\[Prop Gamm 0 compaCT\]:**]{} The Proof is similar as that of propoSiTIon \[PROp Weak cOnvergeNCe\] AND eaSiEr. the PrOOf Will bE omItted. AcKnowledgmeNt {#aCkNowledgment.uNNuMbered} ============== This WorK was supporteD by nationAl natURal ScIenCe FouNdatioN Of ChinA (NSFC) (No. 11431014, No. 11401557), AnD The FunDAmental rESeArch funds for the CentraL unIVersities (No. 0010000048). [0]{} U. BaIer and m.I. fReIDLin, theOrems on larGe DeviaTIons and STaBILIty UNder random perTurbations. [*DOKl. AKad. NauK UsSR*]{} [**235**]{}(1977) 253–256. a. budhirAja anD P. dUpuIs, A variatioNal rEpresentaTion foR PositivE FunctioNals of InfIniTe diMEnSiOnaL BROwnIAn MotIOn. [*PRobab. MatH. STaTist.*]{} [**20**]{}(2000) 39–61. A. budhIRAJA, P. DuPuiS and a. GangUly, Moderate DeViaTion pRinCiple For StOchaStIc DifFerentIal EqUaTions with Jump, arxiv:1401.73v1. x. Chen, ProbAbiLiTieS oF modeRAte devIatIonS for B-vaLued indEPenDeNT RAnDom vectors. [*Chinese J. coNTEmPorary MaTh.*]{} [**11**]{}(1990) 381–393. X. CheN, prObABilities Of ModEratE DEviatIons FOr IndependEnt ranDOm VeCtors in A BAnach sPaCe. [*CHinEse J. APPl. PrObab. StAtist.*]{} [**7**]{}(1991) 24–33. X. ChEn, On lOWer bound of the mODerate deviatiONs OF I.i.D. RandOm vAriables in BAnacH SpacE. [*ChiNEsE J. APPl. ProBab. StAtISt.*]{} [**9**]{}(1993) 386–393. i. chueshov and A. Millet, SToChastiC 2D HydRodynamics TypE Systems: WeLL pOsedness And LARgE deviations. [*Appl. math. OPtim.*]{} [**61**]{}(2010) 379–420. A. Dembo ANd O. ZeitoUni, *LaRge deviaTions techNIQues and aPplIcaTioNs*, (2000) SPRInGer-Verlag, New YORK. H. DjElLout and a. GuIllin, MoDerAte DevIatIoNs for MarkOv chains WiTh AtOm stoChastIC processEs And ThEir AppliCAtions. [*stoch. proc. apPl.*]{} [**95**]{}(2)(2001) 203–221. z. donG, J. X

$, which concludes the pro of of thepropo sit ion . 0.3 cm [**Proof of Pr o psit ion\[Prop Gamm 0 compa ct\]: ** ] {} T h eproof is sim i la r asth at of P r op ositi on\[Propweak conve rge nc e\] and easi e r. The proof wi ll be omitte d. Ackno wl edg m ent { #ac knowl edgmen t .unnum bered} == == = ====== = == Thi s wo rk w as supported by N a ti o nal Natural Sc ienceFo u nd a t ion of China (NS FC ) (No . 114310 1 4, N o . 1 1 401557), andthe Fundame n tal Resea rc h F u nds fo r the C e ntr al Universi ties (No. 001 000004 8 ). [0] { } U. B aier a ndM.I . Fr e id li n,Th e ore m sonl arg e deviat io ns andstab i l i t y un der ran dom p erturbations. [* Dokl . Ak ad. N auk U SSR* ]{ } [** 235**] {}(19 77 ) 253–256. A.Budh iraja and P. D upu is , A v a riatio nal re present ation f o r p os i t i ve functionals of in fi n i te dimensi onal B r ow ni a n motion .[*P roba b . Math . St a ti st.*]{}[**20* * ]{ }( 2000) 3 9– 61. A .Bud hir aja,P . Du puis a nd A. Ga nguly , Moderate Devi a tion Principl e f o r S t ocha sti c Different ialE quat ions wi thJ ump,arXiv :1 4 01 . 73v1. X. Chen, Pro ba biliti es of moderate dev iations fo r B -valuedinde p en d ent random vec tors. [*Chinese J. Conte mpora ry Math. *]{} [**1 1 * *]{}(199 0)381 –39 3.X . C hen, Probabil i t iesof modera tedeviati ons fo r i nde pe ndent ran dom vect or sin a Ba nachs pace. [* Ch ine se J. Appl . Proba b. St atis t. *] { } [ **7**]{ } (1 9 9 1) 2 4– 33 . X. Ch en , Onlowe r bo und ofthe moder ate devi at io ns of i .i.d. randomva riables in B ana ch spa c e . [*Chin ese J. Appl. Probab. St a tist.*] {}[**9* *]{} (1993) 38 6–3 93. I . C h ueshov and A . Mil le t,S t ochas t i c2DHy drodynamic s Typ e Sys te ms:Well Po sedness and LargeD evi ations. [*App l.Math . Op tim . *] { } [ ** 6 1** ] { }(2010) 379–420 . A. Demb oa nd O. Zeitou n i,*L arge de viation s tec h niquesand appli cations*, ( 2000 ) Spr inger-Verl ag, NewYork. H. Djell o ut andA.Guilli n, Mo derat e devi a tio ns fo r Mark ov chain s wit hatom Sto chastic processes and t heir a pplic ati ons. [*St och . Pr oc. Appl. *]{} [**95**]{ }(2 )(2 001)203 – 221. Z.D on g,J . X

$, which_concludes the_proof of the proposition. 0.3cm [**Proof_of Propsition\[Prop_Gamm_0 compact\]:**]{}_The_proof is similar_as that of_Proposition \[Prop weak convergence\]_and easier. The_proof_will be omitted. Acknowledgment {#acknowledgment.unnumbered} ============== This work was supported by National Natural Science Foundation of China_(NSFC)_(No. 11431014,_No._11401557),_and the Fundamental Research Funds_for the Central Universities (No._0010000048). [0]{} U. Baier_and M.I. Freidlin, Theorems on large deviations and_stability_under random perturbations._[*Dokl. Akad. Nauk USSR*]{} [**235**]{}(1977) 253–256. A. Budhiraja and P._Dupuis, A variational representation for positive_functionals of infinite_dimensional_Brownian_motion. [*Probab. Math. Statist.*]{}_[**20**]{}(2000) 39–61. A. Budhiraja, P. Dupuis and_A. Ganguly, Moderate Deviation Principle for_Stochastic Differential Equations with Jump, arXiv:1401.73v1. X. Chen,_Probabilities of moderate deviations for B-valued_independent random vectors. [*Chinese J._Contemporary Math.*]{}_[**11**]{}(1990) 381–393. X. Chen, Probabilities_of moderate deviations_for independent_random vectors in_a Banach space. [*Chinese J. Appl._Probab. Statist.*]{} [**7**]{}(1991)_24–33. X. Chen, On lower bound_of_the moderate deviations_of_i.i.d._random variables_in Banach space._[*Chinese_J. Appl._Probab._Statist.*]{} [**9**]{}(1993) 386–393. I. Chueshov and A._Millet,_Stochastic 2D Hydrodynamics Type Systems: Well Posedness_and Large Deviations. [*Appl._Math._Optim.*]{} [**61**]{}(2010) 379–420. A. Dembo_and O. Zeitouni, *Large deviations_techniques and applications*, (2000) Springer-Verlag, New_York. H. Djellout_and A._Guillin, Moderate deviations for Markov chains with atom Stochastic processes and_their applications. [*Stoch. Proc. Appl.*]{} [**95**]{}(2)(2001)_203–221. Z. Dong, J. X

-7865.4, -7870.3. Lines represent object tracks. Points along the lines represent the position at each time. The larger point at the start of each line denotes the birth position, labelled with the birth time. Grey dots represent observations; those circled are associated with an object track, while others are classified as clutter.\[fig:multiple-object-tracking\]](filter1){width="100.00000%"} ![Phase plots for the multiple object tracking example, showing ground truth (top left) and posterior samples (others) with associated log-weights (left to right, top to bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. Points along the lines represent the position at each time. The larger point at the start of each line denotes the birth position, labelled with the birth time. Grey dots represent observations; those circled are associated with an object track, while others are classified as clutter.\[fig:multiple-object-tracking\]](filter2){width="100.00000%"} ![Phase plots for the multiple object tracking example, showing ground truth (top left) and posterior samples (others) with associated log-weights (left to right, top to bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. Points along the lines represent the position at each time. The larger point at the start of each line denotes the birth position, labelled with the birth time. Grey dots represent observations; those circled are associated with an object track, while others are classified as clutter.\[fig:multiple-object-tracking\]](filter3){width="100.00000%"} ![Phase plots for the multiple object tracking example, showing ground truth (top left) and posterior samples (others) with associated log-weights (left to right, top to bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. Points along the lines represent the position at each time. The larger point at the start of each line denotes the birth position, labelled with the birth time. Grey dots represent observations; those circled are associated with an object track, while others are classified as clutter.\[fig:multiple

-7865.4, -7870.3. Lines represent object tracks. period along the line represent the position at each time. The large point at the start of each tune denote the birth position, labelled with the birth prison term. Grey dots represent observation; those circled are associated with an object track, while others are classified as clutter.\[fig: multiple - object - tracking\]](filter1){width="100.00000% " } ! [ Phase plot for the multiple object tracking example, show ground truth (top left) and posterior sample (others) with associated log - weights (left to right, top to bottomland) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. Points along the lines represent the position at each time. The larger point at the start of each line denotes the birth position, labelled with the parturition time. Grey dots map observation; those circled are associated with an object racetrack, while others are classified as clutter.\[fig: multiple - object - tracking\]](filter2){width="100.00000% " } ! [ Phase plots for the multiple object track example, showing ground truth (top leftover) and posterior samples (others) with associated log - weights (left to right, top to buttocks) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. Points along the line represent the position at each time. The large point at the start of each line denotes the parentage position, pronounce with the birth clock time. Grey dots typify observations; those circle are associated with an aim track, while others are classified as clutter.\[fig: multiple - object - tracking\]](filter3){width="100.00000% " } ! [ Phase plot for the multiple object track example, show ground truth (top left) and posterior samples (others) with associate log - weights (leave to right, top to bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. point along the lines represent the position at each time. The larger point at the beginning of each credit line denote the birth position, labelled with the birth time. Grey dots typify observations; those circled are associated with an aim track, while others are relegate as clutter.\[fig: multiple

-7865.4, -7870.3. Pines represent object tvacks. Points alout the nines depreseng the position at each time. Vhe oargee point at the start ow each lije denotws tie birth positioi, labellcb witg the uirth time. Grey dots reprasent observathovs; those circled are associated with wn objevt track, while ojhers wre dlassified as clutter.\[fig:multiple-obnect-trabking\]](filter1){width="100.00000%"} ![Pnase plots for the multiplf obuect tracking examole, showing grotbd truth (top left) and kovterior sajples (others) with associated loe-weigkts (left to rughh, top to botvom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines vvpresend objecy tracks. Pointx anont the lines represent the position at ewch time. Dhz larger point at the srart mf edch uune dehoves the blrti position, mabelled wirh the birth time. Gter dots represeht obsqrdations; those circled are associated wiuh an object track, while othwrs are classified as clutter.\[fyg:multiple-object-tracking\]](filter2){width="100.00000%"} ![Phase plots fos the ouluiijd ohject tracking example, showing ground truth (tjl kent) and posterior samples (otnegs) rith associatgd log-wzjggts (left to right, hop to fotton) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Linrs represent object tracks. Points along rhe lines represenc the positiun aj each time. The larger point ct the start of ewch line sdnotes the birth poxidion, labtuled with the bireh time. Gcey dpts repfesemt obsqrvations; hhose circled are associwted cith dn object hrack, while others are classifixv as clutter.\[fog:kulniple-objeet-tracling\]](filter3){wideh="100.00000%"} ![Phase plots yor the iultiole object trackiig example, srowing ground jruth (top lefv) and poseeriir sqmples (ughers) with asspciated log-weights (oeft to right, top bo bojtkm) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Linew rtpeesent object yrazks. Pliits ajmng the linev reorerrnt tfe posiunon xt esch time. The larger [oinf at the start of ragh line dgnotes thq birth positoon, labelled with uhe bicth tioe. Grry qots represent observations; thkse circlfd wre associatqd wlth wn object crack, while others are classified as cluvter.\[fig:multiple

-7865.4, -7870.3. Lines represent object tracks. Points lines the position each time. The of line denotes the position, labelled with birth time. Grey dots represent observations; circled are associated with an object track, while others are classified as clutter.\[fig:multiple-object-tracking\]](filter1){width="100.00000%"} plots for the multiple object tracking example, showing ground truth (top left) and samples with log-weights to right, top to bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object tracks. Points the lines represent the position at each time. larger point at the of each line denotes the position, with the time. dots observations; those circled associated with an object track, while others are classified as clutter.\[fig:multiple-object-tracking\]](filter2){width="100.00000%"} ![Phase plots for the multiple object example, showing (top left) posterior (others) associated log-weights (left top to bottom) of -7868.9, -7864.6, Lines represent object tracks. Points along the lines the position each time. The larger point at start of each line denotes the birth position, with the birth time. Grey dots represent observations; those circled are associated with an object others are classified as ![Phase plots for multiple tracking showing truth (top and posterior samples (others) with associated log-weights (left to right, top bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines represent object along lines represent the at each time. The point the start of each the position, birth Grey represent observations; those circled associated with an object track, others are classified as

-7865.4, -7870.3. Lines represent object trackS. Points aloNg the LinEs rEpReseNt thE position at eacH Time. the larger point at the staRt of eAcH Line DEnOtes tHe birth POsITIon, LaBeLleD wITh The biRth Time. GreY dots repreSenT oBservations; tHOsE circled arE asSociated with An oBject tRaCk, wHIle otHerS are cLassifIEd as clUtter.\[fig:mUlTIple-obJEct-tracKINg\]](FiltEr1){width="100.00000%"} ![Phase plots FOr THe multiple objeCt tracKiNG eXAMplE, shOwing grounD tRuth (tOP left) anD PoSTERioR Samples (others) With associaTEd lOg-weigHtS (leFT to rigHt, top To BOttOm) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. Lines rePresEnt object Tracks. pOints alONg the liNes repResEnt The pOSiTiOn aT eACh tIMe. the LArgEr point aT tHe Start Of eaCH LINe deNotEs thE birtH position, labeLleD witH The Birth Time. GRey dOtS reprEsent oBservAtIons; those circleD are AssociateD wiTh An oBjEct trACk, whilE otHerS are claSsified AS clUtTER.\[FiG:multiple-object-traCkING\]](fIlter2){widTh="100.00000%"} ![PhasE PlOtS For the muLtIplE objECT tracKing EXaMple, showIng groUNd TrUth (top lEfT) and poStEriOr sAmpleS (OtheRs) with AssociatEd log-WEights (left to riGHt, top to bottom) OF -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. LINEs REpreSenT object tracKs. PoINts aLong THe LinES reprEsent ThE PoSItion at each time. The lArGer poiNt at tHe start of each Line denoteS THE birth poSitiON, lABelled with the bIrth tIme. Grey dotS RepresenT obseRvations; Those circLED are assoCiaTed WitH an OBJeCt track, while oTHErs aRe ClassifIed As cluttEr.\[fIg:mUltIplE-oBject-tracKing\]](filtEr3){WiDtH="100.00000%"} ![PHasE plotS For the muLtIplE oBjeCt traCKing exAmple, ShowInG gROunD truth (tOP lEFT) and PoStErioR saMpLes (otHers) WIth AssociaTed log-weiGhtS (Left To RiGht, top tO bottom) of -7868.9, -7864.6, -7873.6, -7865.4, -7870.3. LineS rEpresent obJeCt tRacks. POINts along The lines represent the posITion at eAch Time. THe laRger point At tHe starT of EAch linE denotEs the BiRth POSitioN, LAbEllEd With the birTH TimE. Grey DoTs rePresent Observations; those cIRclEd are associatEd wIth aN OBjEct TRaCK, whIlE OthERS are classified aS clutter.\[fiG:mULtIple

-7865.4, -7870.3. Lines r epresent o bject tr ack s. Poi ntsalong the line s rep resent the position at each t i me.T he larg er poin t a t the s ta rtof ea ch li nedenotes the birth po si tion, labell e dwith the b irt h time. Grey do ts rep re sen t obse rva tions ; thos e circl ed are as so c iatedw ith ano b je ct t rack, while other s a r e classified a s clut te r .\ [ f ig: mul tiple-obje ct -trac k ing\]]( f il t e r 1){ w idth="100.000 00%"} ![Ph a seplotsfo r t h e mult ipleob j ect tracking e xamp le, showi ng gro u nd trut h (top l eft) a ndpos teri o rsa mpl es (ot h er s)w ith associa te dlog-w eigh t s ( left to rig ht, t op to bottom) of -78 6 8.9 , -78 64.6, -78 73 .6, - 7865.4 , -78 70 .3. Lines repre sent object t rac ks . P oi nts a l ong th e l ine s repre sent th e po si t i o nat each time. Thela r g er point a t thes ta rt of eachli nedeno t e s the bir t hposition , labe l le dwith th ebirthti me. Gr ey do t s re presen t observ ation s ; those circle d are associat e dw i th an o bje ct track, w hile othe rs a r ecla s sifie d ascl u tt e r.\[fig:multiple-ob je ct-tra cking \]](filter2){ width="100 . 0 0 000%"} ![Ph a se plots for themulti ple object tracking exam ple, sho wing grou n d truth ( top le ft) an d po sterior sampl e s (ot he rs) wit h a ssociat edlog -we igh ts (left to right,to pto b ott om) o f -7868.9 ,-78 64 .6, -787 3 .6, -7 865.4 , -7 87 0. 3 . L ines re p re s e nt o bj ec t tr ack s. Poin ts a l ong the li nes repre sen t the p os ition a t each time.Th e larger p oi ntat the s tart ofeach line denotes the b i rth pos iti on, l abel led withthe birth ti m e. Gre y dots repr es ent o bserv a t io ns; t hose circl e d ar e ass oc iate d withan object track, w h ile others are c las sifi e d a s c l ut t er. \[ f ig: m u ltiple-object-t racking\]] (f i lt er3){width = "10 0. 00000%" } ![Ph ase p l ots for the mult iple obje ct tra c k ing example,showingground tr u th (t o pleft) an d post er ior samp les (o t her s) wi th ass oc iatedlog-w ei ghts (le ft to right, top to bot tom) o f -78 68. 9, -7864. 6,- 787 3.6, -786 5.4, -7870.3.Lin esrepre sen t obje ct t r ac ks. Point s al o ng the li n es re p r es ent the pos i t i onat ea cht ime. T he l arger point at th e start of each lin e den ote s the b irth position, la be l l ed withth e birth tim e. Greydo t s rep resent obser vations ; th o se cir cled ar e associa ted w i th an o bj ec t track , wh il e othe rs are clas s i fied as clutter. \[fig : m ultip l e

-7865.4,_-7870.3. Lines_represent object tracks. Points_along the_lines_represent the_position_at each time._The larger point_at the start of_each line denotes_the_birth position, labelled with the birth time. Grey dots represent observations; those circled are_associated_with an_object_track,_while others are classified as_clutter.\[fig:multiple-object-tracking\]](filter1){width="100.00000%"} ![Phase plots for the multiple_object tracking_example, showing ground truth (top left) and posterior_samples_(others) with associated_log-weights (left to right, top to bottom) of -7868.9,_-7864.6, -7873.6, -7865.4, -7870.3. Lines represent_object tracks. Points_along_the_lines represent the position_at each time. The larger point_at the start of each line_denotes the birth position, labelled with the_birth time. Grey dots represent observations;_those circled are associated with_an object_track, while others are classified_as clutter.\[fig:multiple-object-tracking\]](filter2){width="100.00000%"} ![Phase plots_for the_multiple object tracking_example, showing ground truth (top left)_and posterior samples_(others) with associated log-weights (left to_right,_top to bottom)_of_-7868.9,_-7864.6, -7873.6,_-7865.4, -7870.3. Lines_represent_object tracks._Points_along the lines represent the position_at_each time. The larger point at the_start of each line_denotes_the birth position, labelled_with the birth time. Grey_dots represent observations; those circled are_associated with_an object_track, while others are classified as clutter.\[fig:multiple-object-tracking\]](filter3){width="100.00000%"} ![Phase plots for the multiple_object tracking example, showing ground truth_(top left) and posterior_samples (others)_with_associated log-weights (left_to_right, top_to bottom) of -7868.9, -7864.6, -7873.6, -7865.4,_-7870.3. Lines_represent object tracks. Points along the_lines represent the position_at_each time. The larger point at_the start of each line denotes_the birth position, labelled with_the_birth_time. Grey dots represent observations;_those circled are associated with an_object track, while_others are classified as clutter.\[fig:multiple

\] Let $b {\leqslant}\delta k$ and $\lambda / 2 {\leqslant}\mu {\leqslant}2\lambda - 2$. For each $(C,D) \in {\mathcal{B}}(b)$ with $$|C| {\geqslant}(1 + 2\alpha) \mu b,$$ there are at most $$e^{-\alpha b / 4} \binom{\lambda k/2}{k},$$ sets $A \in {\mathcal{T}}(b) \cap {\mathcal{D}}^*(b,\mu)$ such that $A \cap Y(b) \subset D$. Recall that $|Y(b)| = 2^{16} \lambda^2 b + 2$, and that $b {\geqslant}2^{470} \lambda^{27}$, by , and observe that therefore $|Y(b)|^{5/6} {\leqslant}\alpha \mu b / 2$. By Corollary \[cor:M:containers:app\]$(b)$ and our assumption on $|C|$, it follows that $$|D| {\leqslant}|Y(b)| - \frac{(1 + \alpha) \mu b}{2},$$ and therefore if $A \cap Y(b) \subset D$ then $A'$ misses the set $Y(b) \setminus D \subset [\lambda k / 2]$, which has size at least $(1 + \alpha) \mu b / 2$. Hence, using  and , it follows[^4] that there are at most $$\label{eq:choices:Aprime:Cbig} {\lambda k / 2 - (1 + \alpha) \mu b / 2 \choose k - b} \, {\leqslant}\, e^{5 \delta b} \bigg( \frac{\lambda - 2}{\lambda} \bigg)^{(1+\alpha)\mu b/2} \bigg( \frac{2}{\lambda - 2} \bigg)^{b} \binom{\lambda k/2}{k}$$ choices for $A' = A \cap [\lambda k / 2 ]$ such that $A \cap Y(b) \subset D$. Now, choose the set $B = A \setminus [\lambda k / 2]$, using Lemma \[lem:count:Bs\] and Observation \[obs:log\_concavity\] to bound the number of choices. It follows from  and  that the number of sets $A$ is at most $$e^{5\delta

\ ] Let $ b { \leqslant}\delta k$ and $ \lambda / 2 { \leqslant}\mu { \leqslant}2\lambda - 2$. For each $ (C, D) \in { \mathcal{B}}(b)$ with $ $ |C| { \geqslant}(1 + 2\alpha) \mu b,$$ there are at most $ $ e^{-\alpha b / 4 } \binom{\lambda k/2}{k},$$ sic $ A \in { \mathcal{T}}(b) \cap { \mathcal{D}}^*(b,\mu)$ such that $ A \cap Y(b) \subset D$. Recall that $ |Y(b)| = 2^{16 } \lambda^2 b + 2 $, and that $ b { \geqslant}2^{470 } \lambda^{27}$, by  , and note that therefore $ |Y(b)|^{5/6 } { \leqslant}\alpha \mu b / 2$. By Corollary   \[cor: M: containers: app\]$(b)$ and our presumption on $ |C|$, it follows that $ $ |D| { \leqslant}|Y(b)| - \frac{(1 + \alpha) \mu b}{2},$$ and therefore if $ A \cap Y(b) \subset D$ then $ A'$ miss the set $ Y(b) \setminus D \subset [ \lambda k / 2]$, which has size at least $ (1 + \alpha) \mu b / 2$. therefore, use   and  , it follows[^4 ] that there are at about $ $ \label{eq: choices: Aprime: Cbig } { \lambda k / 2 - (1 + \alpha) \mu b / 2 \choose k - b } \, { \leqslant}\, e^{5 \delta b } \bigg (\frac{\lambda - 2}{\lambda } \bigg)^{(1+\alpha)\mu b/2 } \bigg (\frac{2}{\lambda - 2 } \bigg)^{b } \binom{\lambda k/2}{k}$$ choices for $ A' = A \cap [ \lambda k / 2 ] $ such that $ A \cap Y(b) \subset D$. nowadays, choose the set $ B = A \setminus [ \lambda k / 2]$, using Lemma   \[lem: count: Bs\ ] and Observation   \[obs: log\_concavity\ ] to bound the number of choice. It follows from   and   that the number of sets $ A$ is at about $ $ e^{5\delta

\] Leh $b {\leqslant}\delta k$ and $\uambda / 2 {\leqslaur}\mu {\lexslant}2\mambda - 2$. For each $(C,D) \in {\mathcal{B}}(b)$ wivh $$|C| {\geqsoant}(1 + 2\alpha) \mu b,$$ there are at mlst $$e^{-\alpya b / 4} \binom{\lamusa k/2}{k},$$ sccs $A \jk {\matkcel{T}}(b) \cap {\mathcak{D}}^*(b,\mu)$ such that $A \cap Y(b) \sjbdet D$. Recall that $|Y(b)| = 2^{16} \lambda^2 b + 2$, anq that $n {\heqslant}2^{470} \lambdw^{27}$, by , wnd kbserve that therefore $|Y(b)|^{5/6} {\leqslant}\zlpha \ml b / 2$. By Corollaru \[cor:M:containers:app\]$(b)$ and oug asdumption on $|C|$, it flllows that $$|D| {\lqwslant}|Y(b)| - \frxc{(1 + \alpha) \mu b}{2},$$ and thetefore if $A \cap Y(b) \subset D$ then $A'$ mixses the sgc $Y(b) \vetminus D \wubsen [\lambda k / 2]$, which hav size st least $(1 + \alpma) \mu b / 2$. Hence, using  and , it hollows[^4] that there ate at most $$\lcbel{eq:choices:Aprime:Cbug} {\oambdd k / 2 - (1 + \qlpfa) \ju b / 2 \cholse k - b} \, {\leqalant}\, e^{5 \delra b} \bigg( \frac{\lambds - 2}{\oambda} \bigg)^{(1+\allha)\mu f/2} \figg( \frac{2}{\lambda - 2} \bigg)^{b} \binom{\lambda k/2}{k}$$ bhoides for $A' = A \cap [\lambda k / 2 ]$ such that $A \cap Y(b) \subsee D$. Now, choose the set $B = A \setminus [\lambda k / 2]$, ushng Lxmoa \[ltm:gjjbt:Hs\] and Observation \[obs:log\_concavity\] to bound thq nimner of choices. Ib follows from  and  tjay the number ow sets $A$ js at most $$e^{5\delta

\] Let $b {\leqslant}\delta k$ and $\lambda {\leqslant}\mu - 2$. each $(C,D) \in 2\alpha) b,$$ there are most $$e^{-\alpha b 4} \binom{\lambda k/2}{k},$$ sets $A \in \cap {\mathcal{D}}^*(b,\mu)$ such that $A \cap Y(b) \subset D$. Recall that $|Y(b)| = \lambda^2 b + 2$, and that $b {\geqslant}2^{470} \lambda^{27}$, by , and observe therefore {\leqslant}\alpha b 2$. By Corollary \[cor:M:containers:app\]$(b)$ and our assumption on $|C|$, it follows that $$|D| {\leqslant}|Y(b)| - \frac{(1 \alpha) \mu b}{2},$$ and therefore if $A \cap \subset D$ then $A'$ the set $Y(b) \setminus D [\lambda / 2]$, has at $(1 + \alpha) b / 2$. Hence, using and , it follows[^4] that there are at most $$\label{eq:choices:Aprime:Cbig} {\lambda k 2 - \alpha) \mu / \choose - b} \, \delta b} \bigg( \frac{\lambda - 2}{\lambda} \frac{2}{\lambda - 2} \bigg)^{b} \binom{\lambda k/2}{k}$$ choices for = A [\lambda k / 2 ]$ such $A \cap Y(b) \subset D$. Now, choose the $B = A \setminus [\lambda k / 2]$, using Lemma \[lem:count:Bs\] and Observation \[obs:log\_concavity\] to number of choices. It from and that number sets is most $$e^{5\delta

\] Let $b {\leqslant}\delta k$ and $\lambDa / 2 {\leqslant}\Mu {\leqSlaNt}2\lAmBda - 2$. FOr eaCh $(C,D) \in {\mathcal{B}}(B)$ With $$|c| {\geqslant}(1 + 2\alpha) \mu b,$$ there Are at MoST $$e^{-\alPHa B / 4} \binoM{\lambda K/2}{K},$$ sETS $A \iN {\mAtHcaL{T}}(B) \CaP {\mathCal{d}}^*(b,\mu)$ sucH that $A \cap Y(B) \suBsEt D$. Recall thaT $|y(b)| = 2^{16} \Lambda^2 b + 2$, and ThaT $b {\geqslant}2^{470} \laMbdA^{27}$, by , and ObSerVE that TheReforE $|Y(b)|^{5/6} {\leqSLant}\alPha \mu b / 2$. By COrOLlary \[cOR:M:contaINErS:app\]$(B)$ and our assumption ON $|C|$, IT follows that $$|D| {\lEqslanT}|Y(B)| - \FrAC{(1 + \AlpHa) \mU b}{2},$$ and thereFoRe if $A \CAp Y(b) \subSEt d$ THEn $A'$ MIsses the set $Y(b) \Setminus D \suBSet [\Lambda K / 2]$, wHicH Has sizE at leAsT $(1 + \AlpHa) \mu b / 2$. Hence, uSing  And , it follOws[^4] thaT There arE At most $$\lAbel{eq:ChoIceS:AprIMe:cbIg} {\lAmBDa k / 2 - (1 + \ALpHa) \mU B / 2 \chOose k - b} \, {\leQsLaNt}\, e^{5 \deLta b} \BIGG( \Frac{\LamBda - 2}{\lAmbda} \Bigg)^{(1+\alpha)\mu b/2} \bIgg( \Frac{2}{\LAmbDa - 2} \bigG)^{b} \binOm{\laMbDa k/2}{k}$$ cHoices For $A' = A \CaP [\lambda k / 2 ]$ such thaT $A \caP Y(b) \subset d$. NoW, cHooSe The seT $b = A \setmInuS [\laMbda k / 2]$, usIng LemmA \[Lem:CoUNT:bs\] And Observation \[obs:lOg\_CONcAvity\] to bOund thE NuMbER of choicEs. it fOlloWS From  aNd  thAT tHe number Of sets $a$ Is At Most $$e^{5\deLtA

\] Let $b {\leqslant}\delt a k$ and $ \lamb da/ 2 { \leq slan t}\mu {\leqsla n t}2\ lambda - 2$. For each$(C,D )\ in { \ ma thcal {B}}(b) $ w i t h $ $| C| {\ ge q sl ant}( 1 + 2\alph a) \mu b,$ $ t he re are at mo s t$$e^{-\alp hab / 4} \bino m{\ lambda k /2} { k},$$ se ts $A \in { \ mathca l{T}}(b)\c a p {\ma t hcal{D} } ^ *( b,\m u)$ such that $A\ ca p Y(b) \subsetD$. R ec a ll t hat $| Y(b)| = 2^ {1 6} \l a mbda^2b + 2 $ , a n d that $b {\g eqslant}2^{ 4 70} \lamb da ^{2 7 }$, by  , an do bse rve that th eref ore $|Y(b )|^{5/ 6 } {\leq s lant}\a lpha \ mub / 2$. By C oro ll a ry\ [c or: M :co ntainers :a pp \]$(b )$ a n d o ur a ssu mpti on on $|C|$, it fo llo ws t h at$$|D| {\le qsla nt }|Y(b )| - \ frac{ (1 + \alpha) \mub}{2 },$$ andthe re for eif $A \cap Y (b) \s ubset D $ then$ A'$ m i s s es the set $Y(b) \se tm i n us D \subs et [\l a mb da k / 2]$, w hic h ha s sizeat l e as t $(1 +\alpha ) \ mu b / 2$ .Hence, u sin g and , it f ollows [^4] tha t the r e are at most$ $\label{eq:ch o ic e s :A p rime :Cb ig} {\lambd a k/ 2 - (1+ \ alp h a) \m u b / 2 \c h oose k - b} \, {\le qs lant}\ , e^{ 5 \delta b} \ bigg( \fra c { \ lambda - 2}{ \ la m bda} \bigg)^{( 1+\al pha)\mu b/ 2 } \bigg( \fra c{2}{\la mbda - 2} \ bigg)^{b } \ bin om{ \la m b da k/2}{k}$$ ch o i cesfo r $A' = A\cap [\ lam bda k/ 2 ] $ such th at $A \c ap Y (b )\su bsetD $. Now, c hoo se th e set $B = A \set minu s[\ l amb da k /2 ]$ , usin gLe mma\[l em :coun t:Bs \ ] a nd Obse rvation \ [ob s :log \_ co ncavity \] to bound t he number of c hoi ces. I t followsfrom  and  that the num b er of s ets $A$is a t most $$ e^{ 5\delt a

\] Let_$b {\leqslant}\delta_k$ and $\lambda /_2 {\leqslant}\mu_{\leqslant}2\lambda_- 2$._For_each $(C,D) \in_{\mathcal{B}}(b)$ with $$|C|_{\geqslant}(1 + 2\alpha) \mu_b,$$ there are_at_most $$e^{-\alpha b / 4} \binom{\lambda k/2}{k},$$ sets $A \in {\mathcal{T}}(b) \cap {\mathcal{D}}^*(b,\mu)$ such_that_$A \cap_Y(b)_\subset_D$. Recall that $|Y(b)| = 2^{16}_\lambda^2 b + 2$, and_that $b_{\geqslant}2^{470} \lambda^{27}$, by , and observe that therefore $|Y(b)|^{5/6}_{\leqslant}\alpha_\mu b /_2$. By Corollary \[cor:M:containers:app\]$(b)$ and our assumption on $|C|$, it_follows that $$|D| {\leqslant}|Y(b)| - \frac{(1_+ \alpha) \mu_b}{2},$$_and_therefore if $A \cap_Y(b) \subset D$ then $A'$ misses_the set $Y(b) \setminus D \subset_[\lambda k / 2]$, which has size_at least $(1 + \alpha) \mu_b / 2$. Hence, using _and , it_follows[^4] that there are at_most $$\label{eq:choices:Aprime:Cbig} {\lambda k_/ 2_- (1 +_\alpha) \mu b / 2 \choose_k - b}_\, {\leqslant}\, e^{5 \delta b} \bigg(_\frac{\lambda_- 2}{\lambda} \bigg)^{(1+\alpha)\mu_b/2}_\bigg(_\frac{2}{\lambda -_2} \bigg)^{b} \binom{\lambda_k/2}{k}$$_choices for_$A'_= A \cap [\lambda k /_2_]$ such that $A \cap Y(b) \subset_D$. Now, choose the set_$B_= A \setminus [\lambda_k / 2]$, using Lemma \[lem:count:Bs\]_and Observation \[obs:log\_concavity\] to bound the number_of choices._It follows_from  and  that the number of sets $A$ is at most_$$e^{5\delta

(\Gamma_n^{0\pm}\right)}^{2} + 6 {\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}} \pm 6 {\left(\Gamma_n^{1\pm}\right)}^2 \pm \right.\nonumber\\ && \left. \pm 6 {\Gamma_n^{0\pm}}{\Gamma_n^{2\pm}} + 6 {\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} + 2 {\Gamma_n^{0\pm}}{\Gamma_n^{3\pm}} \right ], \label{five}\end{aligned}$$ $$\begin{aligned} {\frac{\partial^4 Z_{n+1}^{\pm}}{\partial h^4}}=&& y^{\pm 1} \left [ {\left(\Gamma_n^{0\pm}\right)}^{2} \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}} + 12 {\left(\Gamma_n^{1\pm}\right)}^2 + \right.\nonumber\\ && +12 {\Gamma_n^{0\pm}}{\Gamma_n^{2\pm}} \pm 24 {\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{3\pm}} + \nonumber\\ && \left. +6{\left(\Gamma_n^{2\pm}\right)}^2 + 8 {\Gamma_n^{1\pm}}{\Gamma_n^{3\pm}} + 2 {\Gamma_n^{0\pm}}{\Gamma_n^{4\pm}}\right ], \label{six}\end{aligned}$$ where we have used notation $h\equiv \beta H$ and $$\begin{aligned} {\Gamma_n^{i\pm}}\equiv {\frac{\partial^i Z_{n}^{+}}{\partial h^i}}{x^{\pm 1}}+ {\frac{\partial^i Z_{n}^{-}}{\partial h^i}}{x^{\mp 1}}.\end{aligned}$$ Starting from a single spin (0-th generation branch), for which we have ${{\partial^k Z_{0}^{\pm}}/{\partial h^k}}=(\pm 1)^k y^{\pm 1}$, it is straightforward to show by mathematical induction,

( \Gamma_n^{0\pm}\right)}^{2 } + 6 { \Gamma_n^{0\pm}}{\Gamma_n^{1\pm } } \pm 6 { \left(\Gamma_n^{1\pm}\right)}^2 \pm \right.\nonumber\\ & & \left. \pm 6 { \Gamma_n^{0\pm}}{\Gamma_n^{2\pm } } + 6 { \Gamma_n^{1\pm}}{\Gamma_n^{2\pm } } + 2 { \Gamma_n^{0\pm}}{\Gamma_n^{3\pm } } \right ], \label{five}\end{aligned}$$ $ $ \begin{aligned } { \frac{\partial^4 Z_{n+1}^{\pm}}{\partial h^4}}= & & y^{\pm 1 } \left [ { \left(\Gamma_n^{0\pm}\right)}^{2 } \pm 8 { \Gamma_n^{0\pm}}{\Gamma_n^{1\pm } } + 12 { \left(\Gamma_n^{1\pm}\right)}^2 + \right.\nonumber\\ & & +12 { \Gamma_n^{0\pm}}{\Gamma_n^{2\pm } } \pm 24 { \Gamma_n^{1\pm}}{\Gamma_n^{2\pm } } \pm 8 { \Gamma_n^{0\pm}}{\Gamma_n^{3\pm } } + \nonumber\\ & & \left. +6{\left(\Gamma_n^{2\pm}\right)}^2 + 8 { \Gamma_n^{1\pm}}{\Gamma_n^{3\pm } } + 2 { \Gamma_n^{0\pm}}{\Gamma_n^{4\pm}}\right ], \label{six}\end{aligned}$$ where we have used notation $ h\equiv \beta H$ and $ $ \begin{aligned } { \Gamma_n^{i\pm}}\equiv { \frac{\partial^i Z_{n}^{+}}{\partial h^i}}{x^{\pm 1}}+ { \frac{\partial^i Z_{n}^{-}}{\partial h^i}}{x^{\mp 1}}.\end{aligned}$$ Starting from a single tailspin (0 - thursday generation branch), for which we get $ { { \partial^k Z_{0}^{\pm}}/{\partial h^k}}=(\pm 1)^k y^{\pm 1}$, it is aboveboard to show by mathematical trigger,

(\Gamla_n^{0\pm}\right)}^{2} + 6 {\Gamma_n^{0\pm}}{\Gammx_n^{1\pm}} \pm 6 {\left(\Gcnma_n^{1\pm}\cight)}^2 \pj \right.\nunumber\\ && \left. \pm 6 {\Gamma_n^{0\pm}}{\Gamla_b^{2\pm}} + 6 {\Tamma_n^{1\pm}}{\Gamma_n^{2\pm}} + 2 {\Gammx_n^{0\pm}}{\Gamma_j^{3\pm}} \right ], \labtl{five}\end{aligned}$$ $$\usgin{alinued} {\frzg{\partnao^4 Z_{n+1}^{\pm}}{\partial m^4}}=&& y^{\pm 1} \left [ {\left(\Gamma_n^{0\pm}\rhgft)}^{2} \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}} + 12 {\left(\Gamma_n^{1\pm}\ryght)}^2 + \rogjt.\nonumber\\ && +12 {\Gaima_n^{0\km}}{\Gwmma_h^{2\im}} \pm 24 {\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} \pm 8 {\Gammz_n^{0\pm}}{\Gamka_n^{3\pm}} + \nonumbrr\\ && \left. +6{\left(\Gamma_n^{2\pm}\right)}^2 + 8 {\Gwmma_n^{1\pm}}{\Gamma_n^{3\pm}} + 2 {\Gwmma_n^{0\pm}}{\Gammq_n^{4\pm}}\wught ], \label{sib}\end{aligned}$$ where we hzve used notation $h\equiv \beta H$ and $$\yegin{alignee} {\Gqmmw_t^{i\pm}}\equiv {\frec{\partpal^i Z_{n}^{+}}{\partial h^i}}{x^{\pm 1}}+ {\xrac{\paryial^i Z_{n}^{-}}{\partiak h^m}}{x^{\mp 1}}.\end{aligned}$$ Starting fcom a single spin (0-th generatimn branch), for which ww havg ${{\pardial^y Z_{0}^{\po}}/{\padtmal h^k}}=(\pm 1)^n y^{\'m 1}$, it is sfraightforwqrd to show by mathtmaepval inductioh,

(\Gamma_n^{0\pm}\right)}^{2} + 6 {\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}} \pm 6 {\left(\Gamma_n^{1\pm}\right)}^2 && \pm 6 + 6 {\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} \label{five}\end{aligned}$$ {\frac{\partial^4 Z_{n+1}^{\pm}}{\partial h^4}}=&& 1} \left [ \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}} + 12 {\left(\Gamma_n^{1\pm}\right)}^2 \right.\nonumber\\ && +12 {\Gamma_n^{0\pm}}{\Gamma_n^{2\pm}} \pm 24 {\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{3\pm}} + \nonumber\\ && +6{\left(\Gamma_n^{2\pm}\right)}^2 + 8 {\Gamma_n^{1\pm}}{\Gamma_n^{3\pm}} + 2 {\Gamma_n^{0\pm}}{\Gamma_n^{4\pm}}\right ], \label{six}\end{aligned}$$ where we have used $h\equiv H$ $$\begin{aligned} {\frac{\partial^i Z_{n}^{+}}{\partial h^i}}{x^{\pm 1}}+ {\frac{\partial^i Z_{n}^{-}}{\partial h^i}}{x^{\mp 1}}.\end{aligned}$$ Starting from a single spin (0-th generation branch), which we have ${{\partial^k Z_{0}^{\pm}}/{\partial h^k}}=(\pm 1)^k y^{\pm it is straightforward to by mathematical induction,

(\Gamma_n^{0\pm}\right)}^{2} + 6 {\Gamma_n^{0\pm}}{\GammA_n^{1\pm}} \pm 6 {\left(\gamma_N^{1\pm}\RigHt)}^2 \Pm \riGht.\nOnumber\\ && \left. \pm 6 {\GAMma_n^{0\Pm}}{\Gamma_n^{2\pm}} + 6 {\Gamma_n^{1\pm}}{\GammA_n^{2\pm}} + 2 {\GAmMA_n^{0\pm}}{\gAmMa_n^{3\pm}} \Right ], \laBEl{FIVe}\eNd{AlIgnEd}$$ $$\BEgIn{aliGneD} {\frac{\paRtial^4 Z_{n+1}^{\pm}}{\pArtIaL h^4}}=&& y^{\pm 1} \left [ {\lefT(\gaMma_n^{0\pm}\righT)}^{2} \pm 8 {\gamma_n^{0\pm}}{\GammA_n^{1\pM}} + 12 {\left(\GAmMa_n^{1\PM}\righT)}^2 + \riGht.\noNumber\\ && +12 {\gAmma_n^{0\pM}}{\Gamma_n^{2\pm}} \Pm 24 {\gAmma_n^{1\pM}}{\gamma_n^{2\pM}} \PM 8 {\GAmma_N^{0\pm}}{\Gamma_n^{3\pm}} + \nonumbER\\ && \lEFt. +6{\left(\Gamma_n^{2\pm}\Right)}^2 + 8 {\GAmMA_n^{1\PM}}{\gamMa_n^{3\Pm}} + 2 {\Gamma_n^{0\pm}}{\gaMma_n^{4\pM}}\Right ], \laBEl{SIX}\End{ALigned}$$ where we Have used notATioN $h\equiV \bEta h$ And $$\begIn{aliGnED} {\GaMma_n^{i\pm}}\equiV {\fraC{\partial^i z_{n}^{+}}{\partIAl h^i}}{x^{\pm 1}}+ {\FRac{\partIal^i Z_{n}^{-}}{\ParTiaL h^i}}{x^{\MP 1}}.\eNd{AliGnED}$$ StARtIng FRom A single sPiN (0-tH geneRatiON BRAnch), For WhicH we haVe ${{\partial^k Z_{0}^{\pm}}/{\ParTial H^K}}=(\pm 1)^K y^{\pm 1}$, iT is stRaigHtForwaRd to shOw by mAtHematical inductIon,

(\Gamma_n^{0\pm}\right)}^{ 2} + 6 {\G amma_ n^{ 0\p m} }{\G amma _n^{1\pm}} \p m 6{\left(\Gamma_n^{1\pm} \righ t) } ^2 \ p m\righ t.\nonu m be r \ \ & &\l eft .\ pm 6 {\ Gam ma_n^{0 \pm}}{\Gam ma_ n^ {2\pm}} + 6{ \G amma_n^{1\ pm} }{\Gamma_n^{ 2\p m}} +2{\G a mma_n ^{0 \pm}} {\Gamm a _n^{3\ pm}} \rig ht ], \la b el{five } \ en d{al igned}$$ $$\begin { al i gned} {\frac{\ partia l^ 4 Z _ { n+1 }^{ \pm}}{\par ti al h^ 4 }}=&& y ^ {\ p m 1}\ left [ {\lef t(\Gamma_n^ { 0\p m}\rig ht )}^ { 2} \pm 8 {\ Ga m ma_ n^{0\pm}}{\ Gamm a_n^{1\pm }} + 1 2 {\left ( \Gamma_ n^{1\p m}\ rig ht)} ^ 2+\ri gh t .\n o nu mbe r \\&& +12 { \G am ma_n^ {0\p m } } { \Gam ma_ n^{2 \pm}} \pm 24 {\Ga mma _n^{ 1 \pm }}{\G amma_ n^{2 \p m}} \ pm 8 { \Gamm a_ n^{0\pm}}{\Gamm a_n^ {3\pm}} + \n on umb er \\ & & \left . + 6{\ left(\G amma_n^ { 2\p m} \ r i gh t)}^2 + 8 {\Gamma _n ^ { 1\ pm}}{\Ga mma_n^ { 3\ pm } } + 2 {\ Ga mma _n^{ 0 \ pm}}{ \Gam m a_ n^{4\pm} }\righ t ] ,\label{ si x}\end {a lig ned }$$ w h erewe hav e used n otati o n $h\equiv \be t a H$ and $$\b e gi n { al i gned } { \Gamma_n^{i \pm} } \equ iv { \ fr ac{ \ parti al^iZ_ { n} ^ {+}}{\partial h^i}} {x ^{\pm1}}+{\frac{\parti al^i Z_{n} ^ { - }}{\part ialh ^i } }{x^{\mp 1}}.\ end{a ligned}$$Starting from a singl e spin (0 - t h genera tio n b ran ch) , fo r which we ha v e ${{ \p artial^ k Z _{0}^{\ pm} }/{ \pa rti al h^k}}=(\ pm 1)^ky^ {\ pm 1 }$, it i s straigh tf orw ar d t o sho w by ma thema tica lin d uct ion,

(\Gamma_n^{0\pm}\right)}^{2} + 6_{\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}} _\pm 6 {\left(\Gamma_n^{1\pm}\right)}^2 \pm_\right.\nonumber\\ && \left._\pm_6 {\Gamma_n^{0\pm}}{\Gamma_n^{2\pm}}_+ 6_{\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} + 2 {\Gamma_n^{0\pm}}{\Gamma_n^{3\pm}} \right_], \label{five}\end{aligned}$$ $$\begin{aligned} {\frac{\partial^4 Z_{n+1}^{\pm}}{\partial_h^4}}=&& y^{\pm 1} \left [_ {\left(\Gamma_n^{0\pm}\right)}^{2} \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{1\pm}}_+ 12_{\left(\Gamma_n^{1\pm}\right)}^2 + \right.\nonumber\\ && +12 {\Gamma_n^{0\pm}}{\Gamma_n^{2\pm}} \pm 24 {\Gamma_n^{1\pm}}{\Gamma_n^{2\pm}} \pm 8 {\Gamma_n^{0\pm}}{\Gamma_n^{3\pm}} + \nonumber\\ && \left._+6{\left(\Gamma_n^{2\pm}\right)}^2_+ 8_{\Gamma_n^{1\pm}}{\Gamma_n^{3\pm}}_+ 2_{\Gamma_n^{0\pm}}{\Gamma_n^{4\pm}}\right ], \label{six}\end{aligned}$$ where we have_used notation $h\equiv \beta H$_and $$\begin{aligned} {\Gamma_n^{i\pm}}\equiv {\frac{\partial^i_Z_{n}^{+}}{\partial h^i}}{x^{\pm 1}}+ {\frac{\partial^i Z_{n}^{-}}{\partial h^i}}{x^{\mp 1}}.\end{aligned}$$ Starting from a_single_spin (0-th generation_branch), for which we have ${{\partial^k Z_{0}^{\pm}}/{\partial h^k}}=(\pm 1)^k_y^{\pm 1}$, it is straightforward to_show by mathematical_induction,

_{\rm SM} \end{array} \right)=\left( \begin{array}{cc} \cos\theta & -\sin\theta\\ \sin\theta & \cos\theta \end{array} \right)\left( \begin{array}{c} \tilde{h}\\ h' \end{array} \right).$$ Correspondingly, the masses in terms of the input parameters can be written as: $$\begin{aligned} \nonumber m_{H_{\rm light}}^2&=&\lambda_1v^2+\lambda_2 u^2-\sqrt{(\lambda_1 v^2- \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2}, \label{mH1}\\ m_{h_{\rm SM}}^2 &=& \lambda_1v^2+\lambda_2 u^2+\sqrt{(\lambda_1 v^2- \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2}\label{mH2}. \end{aligned}$$ We recall that we identify $h_{\rm SM}$ to be the Higgs discovered at LHC (with $m_{h_{\rm SM}}=125.7$ GeV) and $m_{H_{\rm light}}<m_{h_{\rm SM}}$. Therefore, we need to choose large $\sin\theta$ limit ($\sin\theta\rightarrow 1$) for $h_{\rm SM}$ to get dominant contribution from the SM scalar doublet $H$ whereas $H_{\rm light}$ remains dominantly a scalar singlet (as can be seen from Eq. (\[phi-low\])). From the above mentioned expressions, we can recast the couplings in the scalar potential in terms of the physical quantities like masses and the mixing angels as $$\begin{aligned} \lambda_1 &=& \frac{m_{H_{\rm light}}^2}{4v^2}(1+\cos 2\theta) +\label{l1low} \frac{m_{h_{\rm SM}}^2}{4v^2}(1-\cos 2\theta)\label{l1},\\ \lambda_2 &=& \frac{m_{H_{\rm light}}^2}{4u^2}(1-\cos 2\theta) +

_ { \rm SM } \end{array } \right)=\left ( \begin{array}{cc } \cos\theta & -\sin\theta\\ \sin\theta & \cos\theta \end{array } \right)\left ( \begin{array}{c } \tilde{h}\\ h' \end{array } \right).$$ Correspondingly, the masses in terms of the input argument can be write as: $ $ \begin{aligned } \nonumber m_{H_{\rm light}}^2&=&\lambda_1v^2+\lambda_2 u^2-\sqrt{(\lambda_1 v^2- \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2 }, \label{mH1}\\ m_{h_{\rm SM}}^2 & = & \lambda_1v^2+\lambda_2 u^2+\sqrt{(\lambda_1 v^2- \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2}\label{mH2 }. \end{aligned}$$ We recall that we name $ h_{\rm SM}$ to be the Higgs discovered at LHC (with $ m_{h_{\rm SM}}=125.7 $ GeV) and $ m_{H_{\rm light}}<m_{h_{\rm SM}}$. Therefore, we want to choose large $ \sin\theta$ limit ($ \sin\theta\rightarrow 1 $) for $ h_{\rm SM}$ to get down dominant contribution from the SM scalar doublet $ H$ whereas $ H_{\rm light}$ remains dominantly a scalar singlet (as can be see from Eq. (\[phi - low\ ]) ). From the above mentioned expressions, we can recast the coupling in the scalar potential in terms of the forcible quantities like masses and the mix angels as $ $ \begin{aligned } \lambda_1 & = & \frac{m_{H_{\rm light}}^2}{4v^2}(1+\cos 2\theta) + \label{l1low } \frac{m_{h_{\rm SM}}^2}{4v^2}(1-\cos 2\theta)\label{l1},\\ \lambda_2 & = & \frac{m_{H_{\rm light}}^2}{4u^2}(1-\cos 2\theta) +

_{\rm DM} \end{xrray} \rijht)=\left( \begii{arrqy}{cc} \cus\theta & -\sin\thera\\ \sin\fmeta & \cos\theta \end{array} \right)\left( \nqgin{zgrcy}{c} \tilde{h}\\ h' \end{array} \right).$$ Corrfspondingly, the nasses in tefms of the input paramgters can be written as: $$\begin{aliened} \uonumber m_{H_{\tm llcht}}^2&=&\lambda_1v^2+\lembda_2 l^2-\sqrt{(\lambda_1 v^2- \lakbde_2 u^2)^2+\lqmbda_{12}^2u^2v^2}, \label{mH1}\\ m_{h_{\rm VM}}^2 &=& \lambda_1v^2+\lambda_2 u^2+\sqrt{(\lamtdc_1 v^2- \ldmbdd_2 u^2)^2+\lxnbdx_{12}^2u^2v^2}\mauel{jH2}. \end{wlijned}$$ We reczll that we identify $h_{\rm SM}$ to bq the Higgs diacoverqd at LHC (with $m_{h_{\rm SM}}=125.7$ GeV) and $m_{H_{\rm light}}<k_{h_{\rj SM}}$. Therefore, we need ro choose large $\sin\thgta$ limit ($\fin\theta\rightarrow 1$) for $h_{\rm SM}$ to get dominant cottribntkon fvom gye SM scalar doublet $H$ whereas $H_{\rm light}$ remainf dpmpnantly a scalar finglet (as vaj ng seen from Eq. (\[phi-loc\])). Fdom the above mentloned evpreswions, we san tecast the couplings in the scalar poteutiql in terms of the physical qbantitoes loke masses and the mixiug angsls as $$\begij{aligned} \mxmbda_1 &=& \frac{m_{H_{\rm uigmt}}^2}{4v^2}(1+\wos 2\theta) +\label{l1low} \fcac{m_{h_{\xm SM}}^2}{4v^2}(1-\cor 2\thgta)\labej{l1},\\ \lambda_2 &=& \frac{m_{H_{\rm light}}^2}{4u^2}(1-\cos 2\thetw) +

_{\rm SM} \end{array} \right)=\left( \begin{array}{cc} \cos\theta & & \end{array} \right)\left( \tilde{h}\\ h' \end{array} terms the input parameters be written as: \nonumber m_{H_{\rm light}}^2&=&\lambda_1v^2+\lambda_2 u^2-\sqrt{(\lambda_1 v^2- \lambda_2 \label{mH1}\\ m_{h_{\rm SM}}^2 &=& \lambda_1v^2+\lambda_2 u^2+\sqrt{(\lambda_1 v^2- \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2}\label{mH2}. \end{aligned}$$ We recall that identify $h_{\rm SM}$ to be the Higgs discovered at LHC (with $m_{h_{\rm SM}}=125.7$ and light}}<m_{h_{\rm Therefore, need to choose large $\sin\theta$ limit ($\sin\theta\rightarrow 1$) for $h_{\rm SM}$ to get dominant contribution from SM scalar doublet $H$ whereas $H_{\rm light}$ remains a scalar singlet (as be seen from Eq. (\[phi-low\])). the mentioned expressions, can the in the scalar in terms of the physical quantities like masses and the mixing angels as $$\begin{aligned} \lambda_1 &=& \frac{m_{H_{\rm 2\theta) +\label{l1low} 2\theta)\label{l1},\\ \lambda_2 \frac{m_{H_{\rm 2\theta)

_{\rm SM} \end{array} \right)=\left( \begiN{array}{cc} \coS\thetA & -\siN\thEtA\\ \sin\ThetA & \cos\theta \end{arRAy} \riGht)\left( \begin{array}{c} \tildE{h}\\ h' \enD{aRRay} \rIGhT).$$ CorrEspondiNGlY, THe mAsSeS in TeRMs Of the InpUt paramEters can be WriTtEn as: $$\begin{aliGNeD} \nonumber m_{h_{\rm Light}}^2&=&\lambda_1v^2+\LamBda_2 u^2-\sqRt{(\LamBDa_1 v^2- \laMbdA_2 u^2)^2+\lamBda_{12}^2u^2v^2}, \lABel{mH1}\\ m_{H_{\rm SM}}^2 &=& \lambDa_1V^2+\Lambda_2 U^2+\Sqrt{(\lamBDA_1 v^2- \LambDa_2 u^2)^2+\lambda_{12}^2u^2v^2}\label{mh2}. \EnD{Aligned}$$ We recalL that wE iDEnTIFy $h_{\Rm Sm}$ to be the HiGgS discOVered at lhC (WITH $m_{h_{\RM SM}}=125.7$ GeV) and $m_{H_{\rm Light}}<m_{h_{\rm SM}}$. tHerEfore, wE nEed TO choosE largE $\sIN\thEta$ limit ($\sin\ThetA\rightarrOw 1$) for $h_{\RM SM}$ to geT DominanT contrIbuTioN froM ThE Sm scAlAR doUBlEt $H$ WHerEas $H_{\rm liGhT}$ rEmainS domINANTly a ScaLar sIngleT (as can be seen fRom eq. (\[phI-Low\])). from tHe aboVe meNtIoned ExpresSions, We Can recast the couPlinGs in the scAlaR pOteNtIal in TErms of The PhySical quAntitieS LikE mASSEs And the mixing angels As $$\BEGiN{aligned} \Lambda_1 &=& \FRaC{m_{h_{\Rm light}}^2}{4v^2}(1+\CoS 2\thEta) +\lABEl{l1loW} \fraC{M_{h_{\Rm SM}}^2}{4v^2}(1-\cos 2\Theta)\lABeL{l1},\\ \Lambda_2 &=& \fRaC{m_{H_{\rm lIgHt}}^2}{4u^2}(1-\Cos 2\Theta) +

_{\rm SM} \end{ar ray} \ right)=\left( \b e gin{ a rr ay}{c c} \c os\th eta & -\ sin\theta\ \ \sin \th eta & \cos \th eta \end{ar ra y } \r i gh t )\left( \begin {a r ra y } {c} \ tilde{h } \\ h ' \e nd{ a rray} \right) .$$Correspon dingly , the ma s ses intermsofthe inp u tpa ram et e rsc an be wri tten as: $ $\ begin {ali g n e d } \ non umbe r m_{ H_{\rm light} }^2 &=&\ l amb da_1v ^2+\l ambd a_ 2 u^2 -\sqrt {(\la mb da_1 v^2- \ lambd a _2 u^2 )^2 +\l ambda_{ 12}^2u^ 2 v^2 }, \ l ab el{mH1}\\ m_{h_{\ rm S M} }^2 &=& \lamb d a_ 1v ^ 2+\lambd a_ 2 u ^2+\ s q rt{(\ lamb d a_ 1 v^2- \la mb da_ 2 u ^2)^2 + \lam bda_{1 2}^2u^2v ^2}\l a bel{mH2}. \end { aligned}$$ We re c a ll that we identify $ h_{\ r m SM }$ t o b e t h e Hig gs di sc o ve r ed at LHC (with $m_ {h _{\rmSM}}= 125.7$ GeV) a nd $m_{H_{ \ r m light}} <m_{ h _{ \ rm SM}}$. Ther efore , we needt o choose larg e $\sin\ theta$ li m i t ($\sin \th eta \ri ght a r ro w 1$) for $h_ { \ rm S M} $ to ge t d ominant co ntr ibu tio nfrom theSM scala rdo ub le t $ H$ wh e reas $H_ {\ rmli ght }$ re m ains d omina ntly a s c ala r singl e t( a s ca nbe see n f ro m Eq. (\[ p hi- low\])) . From th e a b oveme nt ioned e xpressions, w ecan recast t hecoupli n g s in the scalar potential in te r ms of t hephysi calquantitie s l ike ma sse s and t he mix ing a ng els a s $$\ b e gi n{a li gned} \la m b da_ 1 &=& \ frac {m_{H_{ \rm light}}^2}{4v^ 2 }(1 +\cos 2\theta ) + \lab e l {l 1lo w } \frac{m_{h_{ \rm SM}}^2 }{ 4 v^ 2}(1-\cos2 \th et a)\labe l{l1},\ \ \l a mbda_2&=& \frac {m_{H_{\r mligh t } }^2 }{4u^2}(1- \cos 2\t heta) +

_{\rm SM} _ _ _ __ __ _ _\end{array} _ __ \right)=\left( __ ___ _ _ _ _\begin{array}{cc} _ _ _ _ _\cos\theta__& -\sin\theta\\ _ _ _ _ \sin\theta & _ \cos\theta _ _ _ _ _ _ _ _ __ \end{array} ___ _ __ __ __ _ __ \right)\left( _ _\begin{array}{c} _ _ _ \tilde{h}\\ _ _ _ __ _h' _ _ _ _ _ \end{array} __ _ \right).$$_Correspondingly, the masses in terms_of_the_input parameters can be written_as: $$\begin{aligned} \nonumber m_{H_{\rm light}}^2&=&\lambda_1v^2+\lambda_2 u^2-\sqrt{(\lambda_1_v^2- _ __ __ ___ \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2}, \label{mH1}\\ m_{h_{\rm_SM}}^2 &=& \lambda_1v^2+\lambda_2 u^2+\sqrt{(\lambda_1 v^2- _ _ _ ___ \lambda_2 u^2)^2+\lambda_{12}^2u^2v^2}\label{mH2}. \end{aligned}$$ We recall that_we identify_$h_{\rm SM}$ to_be the Higgs discovered at LHC (with $m_{h_{\rm SM}}=125.7$ GeV)_and $m_{H_{\rm light}}<m_{h_{\rm SM}}$. Therefore, we need_to choose large $\sin\theta$ limit ($\sin\theta\rightarrow 1$) for $h_{\rm SM}$ to_get dominant contribution from the SM scalar doublet_$H$ whereas $H_{\rm light}$ remains dominantly a_scalar singlet (as can_be_seen from Eq. (\[phi-low\])). From the above_mentioned_expressions,_we_can recast the_couplings in the_scalar potential in_terms of_the physical quantities_like masses_and the mixing_angels_as $$\begin{aligned} \lambda_1 &=& \frac{m_{H_{\rm light}}^2}{4v^2}(1+\cos 2\theta) +\label{l1low} _ \frac{m_{h_{\rm SM}}^2}{4v^2}(1-\cos_2\theta)\label{l1},\\ \lambda_2 &=& \frac{m_{H_{\rm light}}^2}{4u^2}(1-\cos_2\theta)_+

non è immediatamente vicino per i laboratori degli S.U., è certamente lontanissimo per noi italiani che disponiamo di bilanci irrisori a favore della ricerca scientifica (paragonabili solo a quelli dei paesi meno progrediti del mondo)” (cf. [@BernardiniLongoPancini1948], p. 92).[^44] As a matter of fact, after the war, most Italian universities focused their researches on cosmic rays, as these were the cheapest cost means to study particle physics. Padua, in particular, together with Rome and Milan, became in those years one of the most active centers working on cosmic ray researches. The Institute of Physics in Padua – which had been planned and built by Bruno Rossi before the war and was at that time particularly well equipped – had not been damaged and it was an important resource for the renaissance of physics in this university. Let us mention for instance that a large 7 tons electromagnet – that Rossi had designed to deviate charged particles – was ready by 1937 (fig. 14).[^45] Rossi had also started the construction of a 1 million volt accelerator, which would have been the first accelerator of this kind in Italy together with the one in Rome at the Istituto Superiore di Sanità; in November 1938, the setting up of the accelerator was already quite ahead but it was stopped because of the war (fig. 15).[^46] Moreover, Rossi had planned and had made a Wilson cloud chamber to detect and photograph the particles’ tracks. He had not been left time enough to use any of these instruments before his forced departure, but his successors were to do so after the war.[^47] ![The electro-magnet that Rossi had made as part of the equipment of the new Padua Physics Institute in 1937 (reproduced from [@Drigo1939], p. 62).[]{data-label="fig:13"}](Fig13R.jpg){width="\textwidth"} ![The high voltage apparatus for the $10^6$ volt accelerator planned by Rossi in Padua, 1937 (reproduced from [@Drigo1939], p. 61).[]{data-label="fig:"}](Fig14R.jpg){width="\textwidth"} In fact, Antonio Rostagni, Rossi’s successor, decided at first, at the end of

non è immediatamente vicino per i laboratori degli S.U., è certamente lontanissimo per noi italiani che disponiamo di bilanci irrisori a favore della ricerca scientifica (paragonabili solo a quelli dei paesi meno progrediti del mondo) ” (cf. [ @BernardiniLongoPancini1948 ], p. 92).[^44 ] As a matter of fact, after the war, most Italian universities focused their researches on cosmic ray, as these were the bum cost means to study atom physics. Padua, in particular, together with Rome and Milan, become in those years one of the most active center working on cosmic ray inquiry. The Institute of Physics in Padua – which had been planned and built by Bruno Rossi before the war and was at that time peculiarly well equipped – had not been damaged and it was an authoritative resource for the renaissance of physics in this university. permit us mention for instance that a large 7 long ton electromagnet – that Rossi had designed to deviate charged particles – was ready by 1937 (fig. 14).[^45 ] Rossi had besides started the structure of a 1 million volt accelerator, which would have been the first accelerator of this kind in Italy together with the one in Rome at the Istituto Superiore di Sanità; in November 1938, the context up of the accelerator was already quite ahead but it was stopped because of the war (fig. 15).[^46 ] Moreover, Rossi had planned and had made a Wilson swarm bedroom to detect and photograph the atom ’ tracks. He had not been left time enough to use any of these instruments before his forced deviation, but his successor were to do so after the war.[^47 ] ! [ The electro - attraction that Rossi had do as character of the equipment of the new Padua Physics Institute in 1937 (reproduced from [ @Drigo1939 ], p. 62).[]{data - label="fig:13"}](Fig13R.jpg){width="\textwidth " } ! [ The high voltage apparatus for the $ 10 ^ 6 $ volt accelerator planned by Rossi in Padua, 1937 (reproduced from [ @Drigo1939 ], p. 61).[]{data - label="fig:"}](Fig14R.jpg){width="\textwidth " } In fact, Antonio Rostagni, Rossi ’s successor, decide at first, at the end of

noj è immediatamente vicino per i laboratoti degli S.U., è csrtamentd lontanissimo per noi italieni xhe dusponiamo di bilanci ifrisori a favore eelle ricerca scientmrica (pavcgonaglli smoo a quelli del paesi menm progrediti dal mlndo)” (cf. [@BernardiniLongoPancini1948], p. 92).[^44] As a mattrr of fact, after the rar, jost Italian universities focused fheir rtsearches on cosmiv rays, as these were the cjeapfst cost means to dtudy partixle [yysics. Padua, in particlnar, togethgr with Rome and Milan, became in thosz years one od tjg most activx centvrs working ok cosmiw ray rrsearches. The Lnstivute of Physics in Padua – which had been plwnned and bbilt by Bruno Rossi bwfire tve wdr ave wxs zt tgat tile 'articularlg well equipped – had not been caiqged and it wzs an ym[ortant resource for the renaissance of phgsics in this universitt. Let us mention for lnstance ehat a large 7 tons electromagnet – that Rossi had gesigiea ti qdciwte charged particles – was ready by 1937 (fig. 14).[^45] Rosfj nac also started the constrichipg of a 1 milliun volc adcelerator, which wluld hade bewn the fiwst sccelerator of this kind in Italy togetker with the one in Rlme at the Nstitujo Suprriore di Sanità; in Noveober 1938, the settijg up of ffe accelerator wxs slseady quite ahead but it wws stoppev beccuse of ghe ear (fid. 15).[^46] Moreoveg, Rossi had planned and jad mcde a Wilson cllud chamber to detect and photojcaph the partocnes’ tracks. Ke had not been lefe time enough jo use anv of tfese instrlments behore his forsed departure, hut his succxssors wewe ti do so aftdf the war.[^47] ![The ekectro-magutt that Rowsi had made as pavt of fhe equipment oy uhw new Padua Phusizs Yndtmtute hn 1937 (reproducad ffom [@Crigo1939], p. 62).[]{data-labeo="wig:13"}](Fog13R.jpg){width="\textwidth"} ![Dhe gigh voltage apparstms for thg $10^6$ volt ascelerator plsnned by Rossi in Kadua, 1937 (re'roducrd srom [@Drigo1939], p. 61).[]{data-label="fig:"}](Fig14R.jlg){width="\tedtwldth"} In fact, Agtonlo Rjstagni, Roxsi’s successor, decided at first, at the xnd of

non è immediatamente vicino per i laboratori è lontanissimo per italiani che disponiamo della scientifica (paragonabili solo quelli dei paesi progrediti del mondo)” (cf. [@BernardiniLongoPancini1948], p. As a matter of fact, after the war, most Italian universities focused their on cosmic rays, as these were the cheapest cost means to study particle Padua, particular, with and Milan, became in those years one of the most active centers working on cosmic ray The Institute of Physics in Padua – which been planned and built Bruno Rossi before the war was that time well – not been damaged it was an important resource for the renaissance of physics in this university. Let us mention for that a tons electromagnet that had to deviate charged was ready by 1937 (fig. 14).[^45] started the construction of a 1 million volt which would been the first accelerator of this in Italy together with the one in Rome the Istituto Superiore di Sanità; in November 1938, the setting up of the accelerator was ahead but it was because of the (fig. Moreover, had and had a Wilson cloud chamber to detect and photograph the particles’ tracks. had not been left time enough to use any of before forced departure, but successors were to do after war.[^47] ![The electro-magnet that made part of new Physics Institute in 1937 from [@Drigo1939], p. 62).[]{data-label="fig:13"}](Fig13R.jpg){width="\textwidth"} ![The voltage apparatus for the Rossi in Padua, 1937 (reproduced from [@Drigo1939], p. In fact, Antonio Rostagni, Rossi’s successor, decided first, at the end of

non è immediatamente vicino pEr i laboratOri deGli s.U., è CeRtamEnte Lontanissimo peR Noi iTaliani che disponiamo di BilanCi IRrisORi A favoRe della RIcERCa sCiEnTifIcA (PaRagonAbiLi solo a Quelli dei pAesI mEno progreditI DeL mondo)” (cf. [@BeRnaRdiniLongoPaNciNi1948], p. 92).[^44] As a MaTteR Of facT, afTer thE war, moST ItaliAn universItIEs focuSEd their RESeArchEs on cosmic rays, as tHEsE Were the cheapesT cost mEaNS tO STudY paRticle physIcS. PaduA, In partiCUlAR, TOgeTHer with Rome anD Milan, becamE In tHose yeArS onE Of the mOst acTiVE ceNters workinG on cOsmic ray rEsearcHEs. The InSTitute oF PhysiCs iN PaDua – wHIcH hAd bEeN PlaNNeD anD BuiLt by BrunO ROsSi befOre tHE WAR and Was At thAt timE particularly WelL equIPpeD – had nOt beeN damAgEd and It was aN impoRtAnt resource for tHe reNaissance Of pHySicS iN this UNiversIty. let Us mentiOn for inSTanCe THAT a Large 7 tons electromaGnET – ThAt Rossi hAd desiGNeD tO Deviate cHaRgeD parTICles – wAs reADy By 1937 (fig. 14).[^45] RosSi had aLSo StArted thE cOnstruCtIon Of a 1 MilliON volT accelErator, whIch woULd have been the fIRst acceleratoR Of THIs KInd iN ItAly together With THe onE in ROMe At tHE IstiTuto SUpERiORe di Sanità; in NovembeR 1938, tHe settIng up Of the acceleraTor was alreADY Quite aheAd buT It WAs stopped becauSe of tHe war (fig. 15).[^46] MoREover, RosSi had Planned aNd had made A wIlson cloUd cHamBer To dETEcT and photograpH THe paRtIcles’ trAckS. He had nOt bEen LefT tiMe Enough to uSe any of tHeSe InStRumEnts bEFore his fOrCed DeParTure, bUT his suCcessOrs wErE tO Do sO after tHE wAR.[^47] ![the eLeCtRo-maGneT tHat RoSsi hAD maDe as parT of the equIpmENt of ThE nEw Padua physics InstitUtE in 1937 (reproduCeD frOm [@DrigO1939], P. 62).[]{Data-labeL="fig:13"}](Fig13R.jpg){width="\textwidTH"} ![The higH voLtage AppaRatus for tHe $10^6$ vOlt accEleRAtor plAnned bY RossI iN PaDUA, 1937 (reprODUcEd fRoM [@Drigo1939], p. 61).[]{datA-LAbeL="fig:"}](FIg14r.jpg){Width="\teXtwidth"} In fact, AntonIO RoStagni, Rossi’s sUccEssoR, DEcIdeD At FIrsT, aT The END of

non è immediatamente vici no per i l abora tor i d eg li S .U., è certamentel onta nissimo per noi italia ni ch ed ispo n ia mo di bilanc i i r r iso ri a fa vo r edella ri cerca s cientifica (p ar agonabili so l oa quelli d eipaesi meno p rog rediti d elm ondo) ” ( cf. [ @Berna r diniLo ngoPancin i1 9 48], p . 92).[^ 4 4 ] Asa matter of fact, af t er the war, mo st Ita li a nu n ive rsi ties focus ed thei r resear c he s o n c o smic rays, as these were the cheap es t c o st mea ns to s t udy particle p hysi cs. Padua , in p a rticula r , toget her wi thRom e an d M il an, b e cam e i n t h ose years o ne o f the mos t a c tive ce nter s wor king on cosmi c r ay r e sea rches . The Ins ti tuteof Phy sicsin Padua – whichhadbeen plan ned a ndbu ilt b y Bruno Ro ssi before the wa r an dw a s a t that time partic ul a r ly well eq uipped –ha d not bee ndam aged a nd it was an importa nt res o ur ce for th erenais sa nce of phys i cs i n this univers ity.L et us mentionf or instance t h at a l a rge7 t ons electro magn e t –that Ro ssi had d esign ed to deviate charged par ti cles – wasready by 1937 (fig. 14) . [ ^ 45] Ross i ha d a l so started the cons truction o f a 1 mil lionvolt acc elerator, w hich wou ldhav e b een t he first accele r a torof this k ind in Ita lytog eth erwi th the on e in Rom eat t he Is titut o Superio re di S ani tà; i n Novem ber 1 938, t he set ting up of t he a cc el erat orwa s alr eady qui te ahea d but itwas stop pe dbecause of the war ( fi g. 15).[^4 6] Mo reover , Rossi ha d planned and had madea Wilson cl oud c hamb er to det ect and p hot o graphthe pa rticl es ’ t r a cks.H e h adno t been lef t tim e eno ug h to use an y of these instrum e nts before his f orc ed d e p ar tur e ,b uthi s su c c essors were todo so afte rt he war.[^47] ![ Th e elect ro-magn et th a t Rossi had made as partof the e qui pment of t he new P adua Phys i cs In s ti tutein1937 ( re pro duced from[ @Dr igo19 39], p .62).[] {data -l abel="fi g:13"}](Fig13R.jpg){wid th="\t extwi dth "} ![The hi g h v oltage ap para tus for th e $ 10^ 6$ vo lta ccele rato r p lan n ed by Ros s i in Padu a ,193 7 (r eproduced f r o m [@ Drigo 193 9 ], p.61). []{data-label="fi g :"}](Fig14R.jp g){w i d th= "\t e xtwi dt h"} In fact,Ant on i o Rostagn i, Rossi’s su ccessor, d e cided at fi rst, a t the e n d o f

non_è immediatamente_vicino per i laboratori_degli S.U.,_è_certamente lontanissimo_per_noi italiani che_disponiamo di bilanci_irrisori a favore della_ricerca scientifica (paragonabili_solo_a quelli dei paesi meno progrediti del mondo)” (cf. [@BernardiniLongoPancini1948], p. 92).[^44] As a matter_of_fact, after_the_war,_most Italian universities focused their_researches on cosmic rays, as_these were_the cheapest cost means to study particle physics._Padua,_in particular, together_with Rome and Milan, became in those years one_of the most active centers working_on cosmic ray_researches._The_Institute of Physics in_Padua – which had been planned_and built by Bruno Rossi before_the war and was at that time_particularly well equipped – had not_been damaged and it was_an important_resource for the renaissance of_physics in this_university. Let_us mention for_instance that a large 7 tons_electromagnet – that_Rossi had designed to deviate charged_particles_– was ready_by_1937_(fig. 14).[^45]_Rossi had also_started_the construction_of_a 1 million volt accelerator, which_would_have been the first accelerator of this_kind in Italy together_with_the one in Rome_at the Istituto Superiore di_Sanità; in November 1938, the setting_up of_the accelerator_was already quite ahead but it was stopped because of the_war (fig. 15).[^46] Moreover, Rossi had_planned and had made_a Wilson_cloud_chamber to detect_and_photograph the_particles’ tracks. He had not been left_time enough_to use any of these instruments_before his forced departure,_but_his successors were to do so_after the war.[^47] ![The electro-magnet that Rossi_had made as part of_the_equipment_of the new Padua Physics_Institute in 1937 (reproduced from [@Drigo1939],_p. 62).[]{data-label="fig:13"}](Fig13R.jpg){width="\textwidth"} ![The high_voltage apparatus for the $10^6$ volt accelerator_planned_by Rossi in Padua, 1937 (reproduced_from_[@Drigo1939], p. 61).[]{data-label="fig:"}](Fig14R.jpg){width="\textwidth"} In fact, Antonio Rostagni,_Rossi’s_successor,_decided at first, at the_end of

tedious. In this case, we can simply assign numbers to system calls and construct a single `switch` statement which, based on a number, calls the appropriate C function. Therefore, we can rely on the `syscall` system-level primitive without substantial loss of generality or portability. The next question to ask is whether a different extension would serve our purpose better – in particular, there is the obvious choice of exposing each syscall separately as a model checker primitive. There are two arguments against this approach. First, it is desirable that the syscall-related machinery is all in one place and not duplicated in both the stand-in operating system and in the model checker. However, in the *virtual* and *replay* modes, this machinery must be part of the stand-in operating system, which suggests that this should be also the case in the *passthrough* mode. Second, the number of system calls is quite large (typically a few hundred functions) and the functions are system-dependent. When the code that is specific to the host operating system resides in the stand-in operating system, it can be ported once and multiple model checkers can benefit. Of course, the stand-in operating system needs to be ported to the model checker in question, but this offers many additional advantages (particularly the virtual mode). Now if we decide that a single universal primitive becomes part of the model checker, we still need to decide the syntax and the semantics of this extension. Since different system calls take different arguments with varying meaning, the primitive itself will clearly need to be variadic. Since one of the main reasons for choosing a single-primitive interface was platform neutrality, the primitive itself should not possess special knowledge about individual syscalls. First of all, it does not know the bit widths of individual arguments (on most systems, some arguments can be 32 bit – for instance file descriptors – and other 64 bit – object sizes, pointers, etc.). This information is crucial to correctly set up the call to `syscall` (the variadic arguments must line up). Moreover, some pointer-type arguments represent variable-sized *input* data (the buffer argument to `write`, for example) and others represent *output* data (the buffer argument to `read`). In both cases, the size of the memory allocated for the variable-sized argument must be known to `vm_syscall`, so that this memory can be correctly copied between the model check

tedious. In this case, we can simply assign number to system call and construct a single 'switch 'affirmation which, based on a number, call the appropriate carbon affair. Therefore, we can rely on the 'syscall 'organization - level primitive without substantial passing of generalization or portability. The next question to necessitate is whether a different elongation would service our purpose better – in particular, there exist the obvious choice of exposing each syscall separately as a model checker primitive. There be two arguments against this approach. First, it is desirable that the syscall - related machinery is all in one place and not duplicate in both the stand - in operating system and in the model checker. However, in the * virtual * and * replay * modes, this machinery must be part of the stand - in operating system, which suggests that this should be also the case in the * passthrough * mode. Second, the number of system calls is quite large (typically a few hundred functions) and the functions are arrangement - dependent. When the code that is specific to the host operating organization rest in the stand - in operating system, it can be ported once and multiple exemplar checkers can benefit. Of course, the stand - in operating arrangement needs to be ported to the model checker in question, but this offers many additional advantage (particularly the virtual mode). Now if we decide that a single cosmopolitan primitive becomes share of the exemplar checker, we still need to decide the syntax and the semantics of this extension. Since different arrangement calls take different argument with varying meaning, the primitive itself will clearly need to be variadic. Since one of the independent reasons for choose a single - primitive interface was chopine neutrality, the primitive itself should not possess special cognition about individual syscalls. First of all, it does not know the piece widths of individual arguments (on most system, some arguments can be 32 bit – for instance file descriptors – and other 64 bit – object sizes, pointers, etc .). This information is crucial to correctly put up the call to 'syscall '(the variadic arguments must trace up). Moreover, some pointer - type arguments represent variable - sized * input * datum (the buffer argument to 'write `, for example) and others represent * output * data (the buffer controversy to 'take `). In both cases, the size of the memory allocated for the variable - sized argument must be known to 'vm_syscall `, so that this memory can be correctly copied between the exemplar check

tefious. In this case, we cak simply assign uymbers to syatem calus and construct a single `swmtch` stattient which, based ov a numbeg, calls tye akpropriate C funcvjon. Thercyore, sc can cely on the `sysgall` system-nevel primitiva dichout substantial loss of generality or poryahility. The nexj quexeion no ask is whether a different estensioi would serve oir purpose better – in partlculwr, there is the obgious choicg of wxposing eacf syscall separately aa a model checker primitive. Thefe arz two argumgurs wcainst this apprjach. First, it is desisable tnat the syscalk-renatwd machinery is all ii one place and not quplicateg nn both the stand-in opeeatinc syvtem qnd in tie jodel fherker. Howeved, in the *vietual* and *replay* mocef, this machinedy muse fe part of the stand-in operating system, whjch suggests that this whould be also the cade in the *passthrough* mode. Second, the number of system callv is sjitt oarge (ryoically a few hundred functions) and the functyknx sre system-depekdent. When the codr hhsj is specific jo the kksf operating system residef in rhe stand-yn olerating system, it can be pirted once aud nultiple model cheekers can beuefit. Pf coirse, the stand-in operatnng syatem needs ho be porfdd to the model zhebker in quesukon, but this offews many avditipnal adxantsges (pwrticularlj the virtual mode). Now if we dgcide dhat a sinhle universal primitive becomes part of the mpdal bhecker, wz stilk need to decyde the syntax and thz semavtics of tgis extxnsion. Since different syvjem calls takx differegt aegumwnts wigf varying meanong, the pgiiitice itself will cleerly heed to be varicbix. Since one of yhe mayn rxasonf for choosinc a rineke-prioitive intevfaze wss platform neutralidy, tge primitive itselg fhould nit possefs special knpwledge about indigiduan sbscallx. Fyrst of all, it does not know tge bit wifthf of individtal qrguments (on most systems, some arguments can be 32 biv – for instance file dgscriptors – and other 64 bit – objecv sizef, pointerv, etc.). This informatiin is crucial to gorrectly set up the czll to `syscwll` (the variadic arguments must line up). Moreover, some pointer-type argumenrs repcefent variabme-siaed *itpbt* bata (thq buhfxr argument to `wrpte`, for example) and others reprerent *out[uc* data (the buffer argument to `rdad`). In both cxses, the size of the memkry allovated for the variable-sized argiment must be known to `vk_syscall`, so tgat thpa oemory can be corrrctly eopied betweeb thc model check

tedious. In this case, we can simply to calls and a single `switch` number, the appropriate C Therefore, we can on the `syscall` system-level primitive without loss of generality or portability. The next question to ask is whether a extension would serve our purpose better – in particular, there is the obvious of each separately a model checker primitive. There are two arguments against this approach. First, it is desirable that syscall-related machinery is all in one place and duplicated in both the operating system and in the checker. in the and modes, machinery must be of the stand-in operating system, which suggests that this should be also the case in the *passthrough* Second, the system calls quite (typically few hundred functions) functions are system-dependent. When the code to the host operating system resides in the operating system, can be ported once and multiple checkers can benefit. Of course, the stand-in operating needs to be ported to the model checker in question, but this offers many additional the virtual mode). Now we decide that single primitive part the model we still need to decide the syntax and the semantics of extension. Since different system calls take different arguments with varying primitive will clearly need be variadic. Since one the reasons for choosing a was neutrality, should possess knowledge about individual syscalls. of all, it does not the bit widths of some arguments can be 32 bit – for file descriptors – and other 64 bit object sizes, pointers, etc.). This information is crucial to correctly set up call to variadic arguments must line up). Moreover, some pointer-type represent variable-sized *input* data buffer argument to `write`, for example) and others represent data buffer argument `read`). In both the size of memory allocated for argument must known `vm_syscall`, memory can be correctly copied between model

tedious. In this case, we can simPly assign nUmberS to SysTeM calLs anD construct a sinGLe `swItch` statement which, baseD on a nUmBEr, caLLs The apPropriaTE C FUNctIoN. THerEfORe, We can RelY on the `sYscall` systEm-lEvEl primitive wIThOut substanTiaL loss of generAliTy or poRtAbiLIty. ThE neXt queStion tO Ask is wHether a diFfERent exTEnsion wOULd ServE our purpose better – IN pARticular, there iS the obViOUs CHOicE of Exposing eaCh SyscaLL separaTElY AS A moDEl checker primItive. There aRE twO argumEnTs aGAinst tHis apPrOAch. first, it is deSiraBle that thE syscaLL-relateD MachineRy is alL in One PlacE AnD nOt dUpLIcaTEd In bOTh tHe stand-iN oPeRatinG sysTEM ANd in The ModeL checKer. However, in tHe *vIrtuAL* anD *replAy* modEs, thIs MachiNery muSt be pArT of the stand-in opEratIng system, WhiCh SugGeSts thAT this sHouLd bE also thE case in THe *pAsSTHRoUgh* mode. Second, the nuMbER Of System caLls is qUItE lARge (typicAlLy a Few hUNDred fUnctIOnS) and the fUnctioNS aRe System-dEpEndent. whEn tHe cOde thAT is sPecifiC to the hoSt opeRAting system resIDes in the stand-IN oPERaTIng sYstEm, it can be poRted ONce aNd muLTiPle MOdel cHeckeRs CAn BEnefit. Of course, the stAnD-in opeRatinG system needs tO be ported tO THE model chEckeR In QUestion, but this OfferS many additIOnal advaNtageS (particuLarly the vIRTual mode). now If wE deCidE THaT a single univeRSAl prImItive beComEs part oF thE moDel CheCkEr, we still Need to deCiDe ThE sYntAx and THe semantIcS of ThIs eXtensIOn. SincE diffErenT sYsTEm cAlls takE DiFFErenT aRgUmenTs wItH varyIng mEAniNg, the prImitive itSelF Will ClEaRly need To be variadic. SInCe one of the MaIn rEasons FOR choosinG a single-primitive interfACe was plAtfOrm neUtraLity, the prImiTive itSelF Should Not posSess sPeCiaL KNowleDGE aBouT iNdividual sYSCalLs. FirSt Of alL, it does Not know the bit widthS Of iNdividual arguMenTs (on MOSt SysTEmS, SomE aRGumENTs can be 32 bit – for inStance file DeSCrIptors – and oTHer 64 BiT – object Sizes, poInterS, Etc.). This InformatiOn is cruciAl To coRREctLy set up the Call to `syScall` (the vARiadiC ArGumenTs mUst linE uP). MoReoveR, some pOIntEr-typE argumEnTs reprEsent VaRiable-siZed *input* data (the buffer arGument To `wriTe`, fOr example) And OTheRs represeNt *ouTput* data (thE buFfeR arguMenT To `reaD`). In bOTh CasES, the sIze oF The memory ALlOcaTED fOr the variabLE-SIzeD arguMenT Must be KnowN to `vm_syscall`, so thaT This memory can bE corRECtlY coPIed bEtWeen the model chEck

tedious. In this case, we can simpl y ass ign nu mb ersto s ystem calls an d con struct a single `switc h` st at e ment wh ich,based o n a n umb er ,cal ls th e app rop riate C function. T he refore, we c a nrely on th e ` syscall` sys tem -level p rim i tivewit houtsubsta n tial l oss of ge ne r alityo r porta b i li ty.The next question to ask is whether a dif fe r en t ext ens ion wouldse rve o u r purpo s eb e t ter – in particul ar, there i s th e obvi ou s c h oice o f exp os i ngeach syscal l se paratelyas a m o del che c ker pri mitive . T her e ar e t wo ar gu m ent s a gai n stthis app ro ac h. F irst , i t isdes irab le th at the syscal l-r elat e d m achin ery i s al lin on e plac e and n ot duplicated i n bo th the st and -i n o pe ratin g syste m a ndin themodel c h eck er . H ow ever, in the *virt ua l * a nd *repl ay* mo d es ,t his mach in ery mus t be pa rt o f t he stand -in op e ra ti ng syst em , whic hsug ges ts th a t th is sho uld be a lso t h e case in the* passthrough*m od e . S econ d,the numberof s y stem cal l sisq uitelarge ( t yp i cally a few hundred f unctio ns) a nd the functi ons are sy s t e m-depend ent. Wh e n the code tha t isspecific t o the hos t ope rating s ystem res i d es in th e s tan d-i n o p e ra ting system,i t can b e porte d o nce and mu lti ple mo de l checker s can be ne fi t. O f c ourse , the sta nd -in o per ating system need s to b ep ort ed to t h em o delch ec kerinqu estio n, b u t t his off ers manyadd i tion al a dvantag es (particula rl y the virt ua l m ode).N ow if we decide that a single u n iversal pr imiti ve b ecomes pa rtof the mo d el che cker,we st il l n e e d tod e ci deth e syntax a n d th e sem an tics of thi s extension. Since dif ferent system ca llst a ke di f fe r ent a r gum e n ts with varying meaning,th e p rimitive i t sel fwill cl early n eed t o be var iadic. Si nce one o fthem a inreasons fo r choosi ng a sing l e-pri m it ive i nte rfacewa s p latfo rm neu t ral ity,the pr im itiveitsel fshould n ot possess special know ledgeabout in dividualsys c all s. Firstof a ll, it doe s n otknowthe bit w idth s o f i n divid uala rguments( on mo s t s ystems, som e a rgu ments ca n be 32 bit – for instance f i le descriptors – a n d ot her 64 b it – object size s,po i n ters, et c. ). This inf ormation i s cruc ial to corre ctly se t up the ca ll t o ` syscall`(th ev ariadic a rg u mentsmust l ine up ). Mor e over , some pointer-typ e arg u m entsr epr esent v ariable - size d *input*data (the b ufferargu mentto `wri te `, for ex am ple) and o t hers repr esent *outpu t* dat a ( the bu ffer a rgume nt t o`re ad`). Inb o th ca se s , t he s ize o fthememory al l ocated f ort he vari ab le- s i zed ar g um e n t must bekno wn to ` vm_syscall ` , so th a t thi s memo ry can be cor r ect ly copied be t w een the m odel chec k

tedious._In this_case, we can simply_assign numbers_to_system calls_and_construct a single_`switch` statement which,_based on a number,_calls the appropriate_C_function. Therefore, we can rely on the `syscall` system-level primitive without substantial loss of generality_or_portability. The_next_question_to ask is whether a_different extension would serve our_purpose better_– in particular, there is the obvious choice_of_exposing each syscall_separately as a model checker primitive. There are two_arguments against this approach. First, it is_desirable that the_syscall-related_machinery_is all in one_place and not duplicated in both_the stand-in operating system and in_the model checker. However, in the *virtual*_and *replay* modes, this machinery must_be part of the stand-in_operating system,_which suggests that this should_be also the_case in_the *passthrough* mode. Second,_the number of system calls is_quite large (typically_a few hundred functions) and the_functions_are system-dependent. When_the_code_that is_specific to the_host_operating system_resides_in the stand-in operating system, it_can_be ported once and multiple model checkers_can benefit. Of course,_the_stand-in operating system needs_to be ported to the_model checker in question, but this_offers many_additional advantages_(particularly the virtual mode). Now if we decide that a single universal_primitive becomes part of the model_checker, we still need_to decide_the_syntax and the_semantics_of this_extension. Since different system calls take different_arguments with_varying meaning, the primitive itself will_clearly need to be_variadic._Since one of the main reasons_for choosing a single-primitive interface was_platform neutrality, the primitive itself_should_not_possess special knowledge about individual_syscalls. First of all, it does_not know the_bit widths of individual arguments (on most_systems,_some arguments can be 32 bit_–_for instance file descriptors – and_other_64_bit – object sizes, pointers,_etc.). This information is crucial to_correctly set up the call to `syscall` (the variadic_arguments must line_up). Moreover, some pointer-type arguments_represent_variable-sized_*input* data (the buffer argument to `write`, for example) and_others represent_*output* data (the_buffer argument to `read`). In both cases, the size of_the memory allocated for the variable-sized argument_must be known to `vm_syscall`, so that this memory can be_correctly copied between the model check

leq 2^{n-1}R(t)+1} \omega(B_{5(1+k)a_n(t)}(x))\leq k+1\}\big)\\ &\supset\Big\{\max_{n=1}^{z(t)}\sup_{|x| \leq 2^{n-1}(R(t)+1)} \omega(B_{b_n r(t)}(x))\leq k\Big\}\cap \Big\{\sup_{|x| \leq 2^{z(t)}(R(t)+1)} \omega(B_{5(1+k)a_0(t)}(x))\leq k+1\Big\}.\end{aligned}$$ The first event on the right-hand side above occurs a.s. infinitely often by, and the second event occurs eventually by. This yields. Truncation of Poisson potentials {#ss:truncation} -------------------------------- In this section, we control the error that occurs when replacing either an attenuated potential $V^{({\mathfrak{K}})}$ as in or the renormalized potential ${\mkern 1.5mu\overline{\mkern-1.5muV\mkern-1.5mu}\mkern 1.5mu}$ by a truncated potential $V^{(a)} = V^{({\mathfrak{K}}_a)}$, where ${\mathfrak{K}}_a(x) = |x|^{-2}\mathbbm{1}_{\{|x| \le a\}}$. We first state an auxiliary result. \[l:decayPoissonInt\] Let $R \mapsto {\mathfrak{K}}(R) \in L^1(\R^d) \cap L^\infty(\R^d)$ satisfy $$\label{e:conddecayPI} C := \limsup_{R\to\infty} \|{\mathfrak{K}}(R)\|_{L^\infty(\R^d)} < \infty \qquad \text{ and } \qquad \limsup_{R \to \infty} \int_{\R^d} \sup_{|x|\le 1} |{\mathfrak{K}}(R)(x-y)| \dd y < \infty.$$ Then $$\label{e:decayPoissonInt} \limsup_{R \to \infty} \frac{\log \log R}{\log R} \max_{|x| \le R} |V^{({\mathfrak{K}}(R

leq 2^{n-1}R(t)+1 } \omega(B_{5(1+k)a_n(t)}(x))\leq k+1\}\big)\\ & \supset\Big\{\max_{n=1}^{z(t)}\sup_{|x| \leq 2^{n-1}(R(t)+1) } \omega(B_{b_n r(t)}(x))\leq k\Big\}\cap \Big\{\sup_{|x| \leq 2^{z(t)}(R(t)+1) } \omega(B_{5(1+k)a_0(t)}(x))\leq k+1\Big\}.\end{aligned}$$ The first event on the right - hand slope above occur a.s.   infinitely often by, and the second consequence occurs eventually by. This output. Truncation of Poisson potentials { # ss: truncation } -------------------------------- In this part, we control the error that occur when replacing either an attenuated electric potential $ V^{({\mathfrak{K}})}$ as in or the renormalized potential $ { \mkern 1.5mu\overline{\mkern-1.5muV\mkern-1.5mu}\mkern 1.5mu}$ by a truncated electric potential $ V^{(a) } = V^{({\mathfrak{K}}_a)}$, where $ { \mathfrak{K}}_a(x) = |x|^{-2}\mathbbm{1}_{\{|x| \le a\}}$. We first submit an accessory result. \[l: decayPoissonInt\ ] Let $ R \mapsto { \mathfrak{K}}(R) \in L^1(\R^d) \cap L^\infty(\R^d)$ satisfy $ $ \label{e: conddecayPI } speed of light: = \limsup_{R\to\infty } \|{\mathfrak{K}}(R)\|_{L^\infty(\R^d) } < \infty \qquad \text { and } \qquad \limsup_{R \to \infty } \int_{\R^d } \sup_{|x|\le 1 } |{\mathfrak{K}}(R)(x - y)| \dd y < \infty.$$ Then $ $ \label{e: decayPoissonInt } \limsup_{R \to \infty } \frac{\log \log R}{\log R } \max_{|x| \le R } |V^{({\mathfrak{K}}(R

leq 2^{n-1}R(t)+1} \omega(B_{5(1+k)a_n(t)}(x))\leq k+1\}\bin)\\ &\supset\Big\{\max_{n=1}^{z(t)}\sup_{|x| \lex 2^{n-1}(R(t)+1)} \kmega(B_{b_n r(t)}(x))\leq k\Big\}\cap \Big\{\sup_{|x| \leq 2^{v(t)}(R(t)+1)} \omegq(B_{5(1+k)a_0(t)}(x))\leq k+1\Big\}.\end{aligndd}$$ The figst event on uhe right-hand sidx above occurs w.s. inyiiitely often by, and the sacond event ocwufs eventually by. This yields. Truncatiog of Poosdon potentials {#ss:tgugcatjon} -------------------------------- In this section, we control the srror tiat occurs when replacing either an attenkatef potential $V^{({\mathfgak{K}})}$ as in ir trw renormalizdd potential ${\mkern 1.5mu\oberline{\mkern-1.5muV\mkern-1.5mu}\mkern 1.5mu}$ cy a cruncated pitwntldl $V^{(a)} = V^{({\matifrak{K}}_w)}$, where ${\mathngak{K}}_a(x) = |x|^{-2}\mathbnm{1}_{\{|x| \le a\}}$. We fivst svate an auxiliary result. \[l:vecayPoissonInt\] Let $T \mapsto {\mdtkfrak{K}}(R) \in L^1(\R^d) \cap L^\ibfry(\R^d)$ vativfy $$\uqbeu{e:cknvdedayPI} C := \lmmsup_{R\to\inffy} \|{\mathfrak{J}}(R)\|_{L^\infty(\R^d)} < \infty \qauwe \text{ and } \qsuad \liistp_{R \to \infty} \int_{\R^d} \sup_{|x|\le 1} |{\mathfrak{K}}(R)(x-y)| \dd y < \infty.$$ Then $$\label{e:dexayPoissonInt} \limsup_{R \jo \infty} \fwac{\log \log R}{\log R} \max_{|x| \le R} |V^{({\mathfrak{K}}(R

leq 2^{n-1}R(t)+1} \omega(B_{5(1+k)a_n(t)}(x))\leq k+1\}\big)\\ &\supset\Big\{\max_{n=1}^{z(t)}\sup_{|x| \leq 2^{n-1}(R(t)+1)} k\Big\}\cap \leq 2^{z(t)}(R(t)+1)} k+1\Big\}.\end{aligned}$$ The first above a.s. infinitely often and the second occurs eventually by. This yields. Truncation Poisson potentials {#ss:truncation} -------------------------------- In this section, we control the error that occurs replacing either an attenuated potential $V^{({\mathfrak{K}})}$ as in or the renormalized potential ${\mkern 1.5mu}$ a potential = V^{({\mathfrak{K}}_a)}$, where ${\mathfrak{K}}_a(x) = |x|^{-2}\mathbbm{1}_{\{|x| \le a\}}$. We first state an auxiliary result. \[l:decayPoissonInt\] Let \mapsto {\mathfrak{K}}(R) \in L^1(\R^d) \cap L^\infty(\R^d)$ satisfy $$\label{e:conddecayPI} := \limsup_{R\to\infty} \|{\mathfrak{K}}(R)\|_{L^\infty(\R^d)} < \qquad \text{ and } \qquad \to \int_{\R^d} \sup_{|x|\le |{\mathfrak{K}}(R)(x-y)| y \infty.$$ Then $$\label{e:decayPoissonInt} \to \infty} \frac{\log \log R}{\log R} \max_{|x| \le R} |V^{({\mathfrak{K}}(R

leq 2^{n-1}R(t)+1} \omega(B_{5(1+k)a_n(t)}(x))\leq k+1\}\big)\\ &\sUpset\Big\{\maX_{n=1}^{z(t)}\sUp_{|x| \Leq 2^{N-1}(R(T)+1)} \omeGa(B_{b_N r(t)}(x))\leq k\Big\}\cap \bIg\{\suP_{|x| \leq 2^{z(t)}(R(t)+1)} \omega(B_{5(1+k)a_0(t)}(x))\leq K+1\Big\}.\eNd{ALignED}$$ THe firSt event ON tHE RigHt-HaNd sIdE AbOve ocCurS a.s. infiNitely ofteN by, AnD the second evENt Occurs evenTuaLly by. This yieLds. truncaTiOn oF poissOn pOtentIals {#ss:TRuncatIon} -------------------------------- In this SeCTion, we COntrol tHE ErRor tHat occurs when replACiNG either an attenUated pOtENtIAL $V^{({\mAthFrak{K}})}$ as in oR tHe renORmalizeD PoTENTiaL ${\Mkern 1.5mu\overliNe{\mkern-1.5muV\mKErn-1.5Mu}\mkerN 1.5mU}$ by A TruncaTed poTeNTiaL $V^{(a)} = V^{({\mathfraK{K}}_a)}$, wHere ${\mathfRak{K}}_a(x) = |X|^{-2}\Mathbbm{1}_{\{|X| \Le a\}}$. We fiRst staTe aN auXiliARy ReSulT. \[l:DEcaYpoIssONInT\] Let $R \mapStO {\mAthfrAk{K}}(R) \IN l^1(\r^D) \cap l^\inFty(\R^D)$ satiSfy $$\label{e:condDecAyPI} c := \LimSup_{R\tO\inftY} \|{\matHfRak{K}}(R)\|_{l^\infty(\r^d)} < \infTy \Qquad \text{ and } \qquAd \liMsup_{R \to \inFty} \InT_{\R^d} \SuP_{|x|\le 1} |{\mAThfrak{k}}(R)(x-Y)| \dd Y < \infty.$$ THen $$\labeL{E:deCaYpOIsSonInt} \limsup_{R \to \infTy} \FRAc{\Log \log R}{\lOg R} \max_{|X| \Le r} |V^{({\MAthfrak{K}}(r

leq 2^{n-1}R(t)+1} \omega (B_{5(1+k) a_n(t )}( x)) \l eq k +1\} \big)\\ &\sups e t\Bi g\{\max_{n=1}^{z(t)}\s up_{| x| \leq 2^ {n-1} (R(t)+1 ) }\ o meg a( B_ {b_ nr (t )}(x) )\l eq k\Bi g\}\cap \B ig\ {\ sup_{|x| \le q 2 ^{z(t)}(R( t)+ 1)} \omega(B _{5 (1+k)a _0 (t) } (x))\ leq k+1\ Big\}. \ end{al igned}$$Th e first event o n th e ri ght-hand side abo v eo ccurs a.s. inf initel yo ft e n by , a nd the sec on d eve n t occur s e v e n tua l ly by. This y ields. Tru n cat ion of P ois s on pot entia ls {#s s:truncatio n} - --------- ------ - ------- - ------- In t his se ctio n ,we co nt r olt he er r orthat occ ur swhenrepl a c i n g ei the r an atte nuated potent ial $V^ { ({\ mathf rak{K }})} $as in or th e ren or malized potenti al $ {\mkern 1 .5m u\ ove rl ine{\ m kern-1 .5m uV\ mkern-1 .5mu}\m k ern 1 . 5 m u} $ by a truncated p ot e n ti al $V^{( a)} =V ^{ ({ \ mathfrak {K }}_ a)}$ , where ${\ m at hfrak{K} }_a(x) =|x |^{-2}\ ma thbbm{ 1} _{\ {|x | \le a\}} $. Wefirst st ate a n auxiliary res u lt. \[l:deca y Po i s so n Int\ ] L et $R \maps to { \ math frak { K} }(R ) \inL^1(\ R^ d )\ cap L^\infty(\R^d)$ s atisfy $$\l abel{e:condde cayPI} C : = \ limsup_{ R\to \ in f ty} \|{\mathfr ak{K} }(R)\|_{L^ \ infty(\R ^d)}< \infty \qquad\ t ext{ and }\qq uad \l i m su p_{R \to \inf t y } \i nt _{\R^d} \s up_{|x| \le 1} |{ \ma th frak{K}}( R)(x-y)| \ dd y < \i nfty. $ $ Then $ $\ lab el {e: decay P oisson Int}\lim su p_ { R \ to \inf t y} \ frac {\ lo g \l ogR} {\log R}\ma x_{|x|\le R} |V ^{( { \mat hf ra k{K}}(R

leq 2^{n-1}R(t)+1}_ \omega(B_{5(1+k)a_n(t)}(x))\leq_k+1\}\big)\\ &\supset\Big\{\max_{n=1}^{z(t)}\sup_{|x| \leq 2^{n-1}(R(t)+1)} \omega(B_{b_n_r(t)}(x))\leq k\Big\}\cap_\Big\{\sup_{|x|_\leq 2^{z(t)}(R(t)+1)}_\omega(B_{5(1+k)a_0(t)}(x))\leq_k+1\Big\}.\end{aligned}$$ The first_event on the_right-hand side above occurs_a.s. infinitely often by,_and_the second event occurs eventually by. This yields. Truncation of Poisson potentials {#ss:truncation} -------------------------------- In this section,_we_control the_error_that_occurs when replacing either an_attenuated potential $V^{({\mathfrak{K}})}$ as in_or the_renormalized potential ${\mkern 1.5mu\overline{\mkern-1.5muV\mkern-1.5mu}\mkern 1.5mu}$ by a truncated_potential_$V^{(a)} = V^{({\mathfrak{K}}_a)}$,_where ${\mathfrak{K}}_a(x) = |x|^{-2}\mathbbm{1}_{\{|x| \le a\}}$. We first state_an auxiliary result. \[l:decayPoissonInt\] Let $R \mapsto_{\mathfrak{K}}(R) \in L^1(\R^d)_\cap_L^\infty(\R^d)$_satisfy $$\label{e:conddecayPI} C := \limsup_{R\to\infty}_\|{\mathfrak{K}}(R)\|_{L^\infty(\R^d)} < \infty \qquad \text{ and_} \qquad \limsup_{R \to \infty} \int_{\R^d} \sup_{|x|\le_1} |{\mathfrak{K}}(R)(x-y)| \dd y < \infty.$$ Then_$$\label{e:decayPoissonInt} \limsup_{R \to \infty} \frac{\log \log R}{\log_R} \max_{|x| \le R}_|V^{({\mathfrak{K}}(R

top square is commutative by Lemma \[L:commdiag1\], the commutativity of the remaining parts is straightforward. Hence, the diagram (\[E:diagwithresandev\]) is commutative, too. Next, observe that all maps in the diagram (\[E:diagwithresandev\]) but $\bar{\imath}^{!}$ and $\operatorname{\mathsf{ev}}_y$ are isomorphisms. By Lemma \[L:basicsonresandev\], the map $\operatorname{\mathsf{ev}}_y$ is an isomorphism if and only if $n \cdot \bigl([x]-[y]\bigr) \ne 0$ in $J(E)$. This proves the second part of this Proposition. Note that from the exact sequence (\[E:resVectBundle\]) we get the induced map $$R:= H^0\bigl(\underline{\operatorname{\mathsf{res}}}_x^{{\mathcal End}({\mathcal{P}})}\bigr): \quad \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr) {\longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$$ sending an element $g \in \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr)$ to $\bigl(\underline{\operatorname{\mathsf{res}}}_x^{\mathcal{P}}\circ g)_x \in \operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$. Clearly, $R(\imath) = 0$, thus we obtain the induced map $$\label{E:mapRbar} \overline{R}: \quad \frac{\displaystyle \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr)}{\displaystyle \langle \imath \rangle} {\longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr).$$ In the above notation, the following statements are true. 1. $\mathsf{Im}(\overline{R}) = \operatorname{\mathfrak{sl}}\bigl({\mathcal{P}}\big|_x\bigr)$. 2. Moreover, the map $\overline{R}: \quad \frac{\displaystyle \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr)}{\displaystyle \langle \imath \rangle} {\longrightarrow}\operatorname{\

top square is commutative by Lemma \[L: commdiag1\ ], the commutativity of the remaining parts is straightforward. therefore, the diagram (\[E: diagwithresandev\ ]) is commutative, excessively. Next, observe that all maps in the diagram (\[E: diagwithresandev\ ]) but $ \bar{\imath}^{!}$ and $ \operatorname{\mathsf{ev}}_y$ are isomorphism. By Lemma \[L: basicsonresandev\ ], the map $ \operatorname{\mathsf{ev}}_y$ is an isomorphism if and entirely if $ n \cdot \bigl([x]-[y]\bigr) \ne 0 $ in $ J(E)$. This proves the second contribution of this Proposition. Note that from the exact succession (\[E: resVectBundle\ ]) we contract the induce map $ $ R:= H^0\bigl(\underline{\operatorname{\mathsf{res}}}_x^{{\mathcal End}({\mathcal{P}})}\bigr ): \quad \operatorname{\mathsf{Hom}}\bigl({\mathcal{P } }, { \mathcal{P}}(x)\bigr) { \longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$$ sending an element $ gravitational constant \in \operatorname{\mathsf{Hom}}\bigl({\mathcal{P } }, { \mathcal{P}}(x)\bigr)$ to $ \bigl(\underline{\operatorname{\mathsf{res}}}_x^{\mathcal{P}}\circ g)_x \in \operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$. Clearly, $ R(\imath) = 0 $, thus we obtain the induced map $ $ \label{E: mapRbar } \overline{R }: \quad \frac{\displaystyle \operatorname{\mathsf{Hom}}\bigl({\mathcal{P } }, { \mathcal{P}}(x)\bigr)}{\displaystyle \langle \imath \rangle } { \longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr).$$ In the above note, the following statements are true. 1. $ \mathsf{Im}(\overline{R }) = \operatorname{\mathfrak{sl}}\bigl({\mathcal{P}}\big|_x\bigr)$. 2. furthermore, the map $ \overline{R }: \quad \frac{\displaystyle \operatorname{\mathsf{Hom}}\bigl({\mathcal{P } }, { \mathcal{P}}(x)\bigr)}{\displaystyle \langle \imath \rangle } { \longrightarrow}\operatorname{\

too square is commutative ny Lemma \[L:commdict1\], the rommutafivity ow the remaining parts is streighrforwqrd. Hence, the diagram (\[D:diagwithgesandev\]) us cinmutative, vko. Next, observs thac ell maps in the diagram (\[E:giagwithresandav\]) bbt $\bar{\imath}^{!}$ and $\operatorname{\mathsf{ev}}_r$ are ixolorphisms. By Lgmma \[K:fasidsonresandev\], the map $\operatorname{\mzthsf{ev}}_j$ is an isomorphixm if and only if $n \cdot \blgl([x]-[j]\bigr) \ne 0$ in $J(E)$. Thls proves tye sqxond part of this Proposition. Note jhat from the exact sequence (\[E:rerVectYundle\]) we ggc thf induced ma' $$R:= H^0\bpgl(\underline{\oivratorndme{\mathxf{res}}}_x^{{\mathcal Cnd}({\mavhcao{P}})}\bigr): \quad \operatoriame{\mathsf{Hom}}\bigl({\matrcal{P}}, {\matvccl{P}}(x)\bigr) {\longrightarriw}\iperajornake{\magysf{Dnd}}\gijl({\mzthcal{O}}\bij|_x\bigr)$$ sendjng an elemwnt $g \in \operatornake{\iqthsf{Hom}}\bigl({\mzthcal{[}}, {\iathcal{P}}(x)\bigr)$ to $\bigl(\underline{\operatorndme{\jathsf{res}}}_x^{\mathcal{P}}\circ t)_x \in \operatorname{\mathdf{End}}\bigl({\iathcal{P}}\big|_x\bigr)$. Clearly, $R(\imath) = 0$, thus we obtain dhe iidjceb map $$\uqbfl{E:mapRbar} \overline{R}: \quad \frac{\displaystyle \opewztprkame{\mathsf{Hom}}\bigj({\mathcal{P}}, {\mstjcsj{P}}(x)\bigr)}{\displaistyle \lanfle \imath \rangle} {\lojgrightwrrow}\iperatornwme{\msthsf{End}}\bigl({\mathcal{P}}\big|_x\bige).$$ In the abovv noration, the followiug statementr arg true. 1. $\mathsf{Im}(\overline{R}) = \opzratorhame{\mathfran{sl}}\bigl({\maffcal{P}}\big|_x\bigr)$. 2. Murepvar, the map $\overline{R}: \quad \srac{\displeystyke \operxtormame{\maehsf{Hom}}\bigp({\mathgdl{P}}, {\mathcal{P}}(x)\bigr)}{\dlsplaistyle \langle \imwth \rangle} {\longrightarrow}\opecetorname{\

top square is commutative by Lemma \[L:commdiag1\], of remaining parts straightforward. Hence, the Next, that all maps the diagram (\[E:diagwithresandev\]) $\bar{\imath}^{!}$ and $\operatorname{\mathsf{ev}}_y$ are isomorphisms. By \[L:basicsonresandev\], the map $\operatorname{\mathsf{ev}}_y$ is an isomorphism if and only if $n \cdot \ne 0$ in $J(E)$. This proves the second part of this Proposition. Note from exact (\[E:resVectBundle\]) get the induced map $$R:= H^0\bigl(\underline{\operatorname{\mathsf{res}}}_x^{{\mathcal End}({\mathcal{P}})}\bigr): \quad \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr) {\longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$$ sending an element $g \in {\mathcal{P}}(x)\bigr)$ to $\bigl(\underline{\operatorname{\mathsf{res}}}_x^{\mathcal{P}}\circ g)_x \in \operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$. Clearly, $R(\imath) 0$, thus we obtain induced map $$\label{E:mapRbar} \overline{R}: \quad \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, \langle \imath {\longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr).$$ the notation, the following are true. 1. $\mathsf{Im}(\overline{R}) = \operatorname{\mathfrak{sl}}\bigl({\mathcal{P}}\big|_x\bigr)$. 2. Moreover, the map $\overline{R}: \quad \frac{\displaystyle \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr)}{\displaystyle \langle \imath {\longrightarrow}\operatorname{\

top square is commutative by LEmma \[L:commdIag1\], thE coMmuTaTiviTy of The remaining paRTs is Straightforward. Hence, thE diagRaM (\[e:diaGWiThresAndev\]) is COmMUTatIvE, tOo. NExT, ObServe ThaT all mapS in the diagRam (\[e:dIagwithresanDEv\]) But $\bar{\imatH}^{!}$ anD $\operatornamE{\maThsf{ev}}_Y$ aRe iSOmorpHisMs. By LEmma \[L:bASicsonResandev\], tHe MAp $\operATorname{\MAThSf{ev}}_Y$ is an isomorphism iF AnD Only if $n \cdot \bigL([x]-[y]\bigR) \nE 0$ In $j(e)$. thiS prOves the secOnD part OF this PrOPoSITIon. nOte that from thE exact sequeNCe (\[E:ResVecTBUndLE\]) we get The inDuCEd mAp $$R:= H^0\bigl(\undErliNe{\operatoRname{\mAThsf{res}}}_X^{{\Mathcal end}({\matHcaL{P}})}\bIgr): \qUAd \OpEraToRNamE{\MaThsF{hom}}\Bigl({\mathCaL{P}}, {\MathcAl{P}}(x)\BIGR) {\LongRigHtarRow}\opEratorname{\matHsf{end}}\bIGl({\mAthcaL{P}}\big|_X\bigR)$$ sEndinG an eleMent $g \In \Operatorname{\matHsf{HOm}}\bigl({\matHcaL{P}}, {\MatHcAl{P}}(x)\bIGr)$ to $\biGl(\uNdeRline{\opEratornAMe{\mAtHSF{ReS}}}_x^{\mathcal{P}}\circ g)_x \in \OpERAtOrname{\maThsf{EnD}}\BiGl({\MAthcal{P}}\bIg|_X\biGr)$. ClEARly, $R(\iMath) = 0$, THuS we obtaiN the inDUcEd Map $$\labeL{E:MapRbaR} \oVerLinE{R}: \quaD \Frac{\DisplaYstyle \opEratoRName{\mathsf{Hom}}\bIGl({\mathcal{P}}, {\matHCaL{p}}(X)\bIGr)}{\diSplAystyle \langLe \imATh \raNgle} {\LOnGriGHtarrOw}\opeRaTOrNAme{\mathsf{End}}\bigl({\matHcAl{P}}\big|_X\bigr).$$ in the above notAtion, the foLLOWing statEmenTS aRE true. 1. $\mathsf{Im}(\oVerliNe{R}) = \operatoRName{\mathFrak{sL}}\bigl({\matHcal{P}}\big|_x\BIGr)$. 2. MoreovEr, tHe mAp $\oVerLINe{r}: \quad \frac{\dispLAYstyLe \OperatoRnaMe{\mathsF{HoM}}\biGl({\mAthCaL{P}}, {\mathcal{p}}(x)\bigr)}{\diSpLaYsTyLe \lAngle \IMath \rangLe} {\LonGrIghTarroW}\OperatOrnamE{\

top square is commutative by Lemma\[L:c omm dia g1 \],thecommutativityo f th e remaining parts is s traig ht f orwa r d. Henc e, thed ia g r am(\ [E :di ag w it hresa nde v\]) is commutati ve, t oo. Next, o b se rve that a llmaps in thedia gram ( \[ E:d i agwit hre sande v\]) b u t $\ba r{\imath} ^{ ! }$ and $\opera t o rn ame{ \mathsf{ev}}_y$ a r ei somorphisms. B y Lemm a\ [L : b asi cso nresandev\ ], them ap $\op e ra t o r nam e {\mathsf{ev}} _y$ is an i s omo rphism i f a n d only if $ n\ cdo t \bigl([x] -[y] \bigr) \n e 0$ i n $J(E)$ . This p rovesthe se cond pa rt of t h isP ro pos i tio n. Note t ha t fro m th e e x actseq uenc e (\[ E:resVectBund le\ ]) w e ge t the indu cedma p $$R := H^0 \bigl (\ underline{\oper ator name{\mat hsf {r es} }} _x^{{ \ mathca l E nd} ({\math cal{P}} ) }\b ig r ) : \quad \operatornam e{ \ m at hsf{Hom} }\bigl ( {\ ma t hcal{P}} ,{\m athc a l {P}}( x)\b i gr ) {\long righta r ro w} \operat or name{\ ma ths f{E nd}}\ b igl( {\math cal{P}}\ big|_ x \bigr)$$ sendi n g an element$ g\ i n\ oper ato rname{\math sf{H o m}}\ bigl ( {\ mat h cal{P }}, { \m a th c al{P}}(x)\bigr)$ to $ \bigl( \unde rline{\operat orname{\ma t h s f{res}}} _x^{ \ ma t hcal{P}}\circg)_x\in \opera t orname{\ maths f{End}}\ bigl({\ma t h cal{P}}\ big |_x \bi gr) $ . C learly, $R(\i m a th)=0$, thu s w e obtai n t heind uce dmap $$\la bel{E:ma pR ba r} \ ove rline { R}: \qua d\fr ac {\d ispla y style\oper ator na me { \ma thsf{Ho m }} \ b igl( {\ ma thca l{P }} , {\m athc a l{P }}(x)\b igr)}{\di spl a ysty le \ langle\imath \rangl e} {\longrig ht arr ow}\op e r atorname {\mathsf{End}}\bigl({\m a thcal{P }}\ big|_ x\bi gr).$$ I n t he abo ven otatio n, the foll ow ing s tatem e n ts ar etrue. 1.$ \ma thsf{ Im }(\o verline {R}) = \operatorna m e{\ mathfrak{sl}} \bi gl({ \ m at hca l {P } }\b ig | _x\ b i gr)$. 2. More over, thema p $ \overline{ R }:\q uad \fr ac{\dis plays t yle \op eratornam e{\mathsf {H om}} \ b igl ({\mathcal {P}}, {\ mathcal{P } }(x)\ b ig r)}{\ dis playst yl e \ langl e \ima t h \ rangl e} { \longr ighta rr ow}\oper atorname{\

top_square is_commutative by Lemma \[L:commdiag1\],_the commutativity_of_the remaining_parts_is straightforward. Hence,_the diagram (\[E:diagwithresandev\])_is commutative, too. Next, observe_that all maps_in_the diagram (\[E:diagwithresandev\]) but $\bar{\imath}^{!}$ and $\operatorname{\mathsf{ev}}_y$ are isomorphisms. By Lemma \[L:basicsonresandev\], the map_$\operatorname{\mathsf{ev}}_y$_is an_isomorphism_if_and only if $n \cdot_\bigl([x]-[y]\bigr) \ne 0$ in $J(E)$._This proves_the second part of this Proposition. Note that from_the_exact sequence (\[E:resVectBundle\])_we get the induced map $$R:= H^0\bigl(\underline{\operatorname{\mathsf{res}}}_x^{{\mathcal End}({\mathcal{P}})}\bigr): _\quad \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr) {\longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$$ sending an_element $g \in_\operatorname{\mathsf{Hom}}\bigl({\mathcal{P}},_{\mathcal{P}}(x)\bigr)$_to $\bigl(\underline{\operatorname{\mathsf{res}}}_x^{\mathcal{P}}\circ g)_x \in \operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr)$._Clearly, $R(\imath) = 0$, thus we_obtain the induced map $$\label{E:mapRbar} \overline{R}: \quad_\frac{\displaystyle \operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr)}{\displaystyle \langle \imath \rangle} {\longrightarrow}\operatorname{\mathsf{End}}\bigl({\mathcal{P}}\big|_x\bigr).$$ In the_above notation, the following statements are_true. 1. $\mathsf{Im}(\overline{R}) = \operatorname{\mathfrak{sl}}\bigl({\mathcal{P}}\big|_x\bigr)$. 2._ Moreover,_the map $\overline{R}: \quad \frac{\displaystyle_\operatorname{\mathsf{Hom}}\bigl({\mathcal{P}}, {\mathcal{P}}(x)\bigr)}{\displaystyle \langle_\imath \rangle} _ _{\longrightarrow}\operatorname{\

}{q})2\pi$ ($r,~q$, co-prime integers, $q>0$), in which case both the amplitude and phase are repeated at intervals of $\tilde{N}=q N$ lattice sites. On the other hand, if $\frac{\delta}{2\pi}$ is irrational, (\[eq:ten\]) represents an [*incommensurate*]{} standing wave since the phase of $\psi_n$ never repeats itself, but comes arbitrarily close to any preassigned value $\bar{\delta}$ for some $n$ or other..5cm In section (5), we consider for the sake of concreteness two particular [*commensurate*]{} solutions with $\delta=\pi$ and with $N=2,~3$ respectively. The solution for $N=2,~\delta=\pi~(\tilde{N}=4)$ reads (up to translation by a lattice site) $$\begin{aligned} {\bar\phi}_0=-{\bar\phi}_2=b\label{eq:thirteena},\\ {\bar\phi}_1={\bar\phi}_3=0 \label{eq:thirteenb},\\ {\bar\phi}_{n+\tilde{N}}={\bar\phi}_n, \label{eq:thirteenc}\end{aligned}$$ where $$b=\frac{\gamma}{\gamma+\omega}, \label{eq:thirteend}$$ and where $\gamma$ is to satisfy the consistency condition $$\gamma>\gamma_1(\lambda)|_{N=2,~\delta=\pi} =-\omega. \label{eq:thirteene}$$ .5cm Recall that in this paper we restrict to $\omega<-2$, while similar results are obtained for $\omega>2$. In a similar manner, one has, for $N=3$, $\delta=\pi$ (periodicity $\tilde{N}=6$) $$\begin{aligned} {\bar\phi}_0=-{\bar\phi}_3=b, \label{eq:fourteena}\\ {\bar\phi}_1=-{\bar\phi}_2=-{\bar\phi}_4={\bar\phi}_5=\frac{b}{1-\omega}, \label{eq:fourteenb}\\ {\bar\phi}_{n+\tilde{N}}={\bar\phi}_n, \label{eq:fourteenc} \

} { q})2\pi$ ($ r,~q$, co - prime integers, $ q>0 $), in which case both the amplitude and phase are repeated at interval of $ \tilde{N}=q N$ wicket sites. On the other bridge player, if $ \frac{\delta}{2\pi}$ is irrational, (\[eq: ten\ ]) typify an [ * incommensurate * ] { } standing wave since the phase of $ \psi_n$ never repeats itself, but come arbitrarily close to any preassigned value $ \bar{\delta}$ for some $ n$ or early.. 5 centimeter In section (5), we regard for the sake of concreteness two particular [ * commensurate * ] { } solutions with $ \delta=\pi$ and with $ N=2,~3 $ respectively. The solution for $ N=2,~\delta=\pi~(\tilde{N}=4)$ read (up to translation by a lattice site) $ $ \begin{aligned } { \bar\phi}_0=-{\bar\phi}_2 = b\label{eq: thirteena},\\ { \bar\phi}_1={\bar\phi}_3=0 \label{eq: thirteenb},\\ { \bar\phi}_{n+\tilde{N}}={\bar\phi}_n, \label{eq: thirteenc}\end{aligned}$$ where $ $ b=\frac{\gamma}{\gamma+\omega }, \label{eq: thirteend}$$ and where $ \gamma$ is to satisfy the consistency condition $ $ \gamma>\gamma_1(\lambda)|_{N=2,~\delta=\pi } = -\omega. \label{eq: thirteene}$$ .5 cm Recall that in this paper we qualify to $ \omega<-2 $, while similar results are obtained for $ \omega>2$. In a like manner, one has, for $ N=3 $, $ \delta=\pi$ (periodicity $ \tilde{N}=6 $) $ $ \begin{aligned } { \bar\phi}_0=-{\bar\phi}_3 = b, \label{eq: fourteena}\\ { \bar\phi}_1=-{\bar\phi}_2=-{\bar\phi}_4={\bar\phi}_5=\frac{b}{1-\omega }, \label{eq: fourteenb}\\ { \bar\phi}_{n+\tilde{N}}={\bar\phi}_n, \label{eq: fourteenc } \

}{q})2\pi$ ($r,~q$, co-prime integers, $q>0$), ik which case botk the akplituse and pfase are repeated at intervaps of $\tulde{N}=q N$ lattice sites. On the onher hand, if $\hrac{\delta}{2\pi}$ is icdational, (\[eq:teh\]) repxewents an [*incomkensurate*]{} vtanding wave vivcz the phase of $\psi_n$ never repeats itfelf, buy fomes arbitraryly bljse fo any preassigned value $\bar{\delta}$ ror somt $n$ or other..5cm In xection (5), we consider for tje swke of concretenesd two partixulaw [*commensuratd*]{} solutions with $\delta=\ki$ and with $N=2,~3$ respectively. The sulutipn for $N=2,~\deotq=\pi~(\hhlde{N}=4)$ reads (up tj translatiok by a nattice site) $$\begin{aligked} {\bac\phi}_0=-{\var\phi}_2=b\label{eq:thirteeia},\\ {\bar\phi}_1={\bar\phi}_3=0 \label{gq:thirteent},\\ {\bcr\phi}_{n+\tilde{N}}={\bar\phi}_n, \lqbwl{eq:tvirtaenc}\dbd{auighev}$$ whsre $$b=\fgac{\jamma}{\gamma+\ojega}, \label{ew:thirteend}$$ and whert $\gwnma$ is to satjsfy tre consistency condition $$\gamma>\gamma_1(\lambda)|_{T=2,~\demta=\pi} =-\omega. \label{eq:thirreene}$$ .5cm Recall that ij this pa[er we restrict to $\omega<-2$, while similar results ara obteived fjf $\olega>2$. In a similar manner, one has, for $N=3$, $\delta=\py$ (ptripdicity $\tilde{N}=6$) $$\negin{aligned} {\bar\phi}_0=-{\nag\pny}_3=b, \label{eq:foutteena}\\ {\bcd\pgi}_1=-{\bar\phi}_2=-{\bar\phi}_4={\bar\pji}_5=\frac{b}{1-\jmega}, \label{eq:fjurtrenb}\\ {\bar\phi}_{n+\tilde{N}}={\bar\phi}_n, \lavel{eq:fourteeuc} \

}{q})2\pi$ ($r,~q$, co-prime integers, $q>0$), in which the and phase repeated at intervals On other hand, if is irrational, (\[eq:ten\]) an [*incommensurate*]{} standing wave since the of $\psi_n$ never repeats itself, but comes arbitrarily close to any preassigned value for some $n$ or other..5cm In section (5), we consider for the sake concreteness particular solutions $\delta=\pi$ and with $N=2,~3$ respectively. The solution for $N=2,~\delta=\pi~(\tilde{N}=4)$ reads (up to translation by a lattice $$\begin{aligned} {\bar\phi}_0=-{\bar\phi}_2=b\label{eq:thirteena},\\ {\bar\phi}_1={\bar\phi}_3=0 \label{eq:thirteenb},\\ {\bar\phi}_{n+\tilde{N}}={\bar\phi}_n, \label{eq:thirteenc}\end{aligned}$$ where $$b=\frac{\gamma}{\gamma+\omega}, and where $\gamma$ is satisfy the consistency condition $$\gamma>\gamma_1(\lambda)|_{N=2,~\delta=\pi} \label{eq:thirteene}$$ Recall that this we to $\omega<-2$, while results are obtained for $\omega>2$. In a similar manner, one has, for $N=3$, $\delta=\pi$ (periodicity $\tilde{N}=6$) $$\begin{aligned} \label{eq:fourteena}\\ {\bar\phi}_1=-{\bar\phi}_2=-{\bar\phi}_4={\bar\phi}_5=\frac{b}{1-\omega}, \label{eq:fourteenc} \

}{q})2\pi$ ($r,~q$, co-prime integers, $q>0$), in whIch case botH the aMplItuDe And pHase Are repeated at iNTervAls of $\tilde{N}=q N$ lattice siTes. On ThE OtheR HaNd, if $\fRac{\deltA}{2\Pi}$ IS IrrAtIoNal, (\[Eq:TEn\]) RepreSenTs an [*incOmmensuratE*]{} stAnDing wave sincE ThE phase of $\psI_n$ nEver repeats iTseLf, but cOmEs aRBitraRilY closE to any PReassiGned value $\BaR{\Delta}$ fOR some $n$ oR OThEr..5cm in section (5), we considER fOR the sake of concReteneSs TWo PARtiCulAr [*commensuRaTe*]{} solUTions wiTH $\dELTA=\pi$ ANd with $N=2,~3$ respecTively. The soLUtiOn for $N=2,~\DeLta=\PI~(\tilde{n}=4)$ readS (uP To tRanslation bY a laTtice site) $$\Begin{aLIgned} {\baR\Phi}_0=-{\bar\pHi}_2=b\labEl{eQ:thIrteENa},\\ {\BaR\phI}_1={\bAR\phI}_3=0 \LaBel{EQ:thIrteenb},\\ {\bAr\PhI}_{n+\tilDe{N}}={\bAR\PHI}_n, \laBel{Eq:thIrteeNc}\end{aligned}$$ wHerE $$b=\frAC{\gaMma}{\gaMma+\omEga}, \lAbEl{eq:tHirteeNd}$$ and WhEre $\gamma$ is to satIsfy The consisTenCy ConDiTion $$\gAMma>\gamMa_1(\lAmbDa)|_{N=2,~\deltA=\pi} =-\omegA. \LabEl{EQ:THiRteene}$$ .5cm Recall that In THIs Paper we rEstricT To $\OmEGa<-2$, while sImIlaR resULTs are ObtaINeD for $\omegA>2$. In a siMIlAr Manner, oNe Has, for $n=3$, $\dEltA=\pi$ (PerioDIcitY $\tilde{n}=6$) $$\begin{alIgned} {\BAr\phi}_0=-{\bar\phi}_3=b, \laBEl{eq:fourteena}\\ {\BAr\PHI}_1=-{\bAR\phi}_2=-{\Bar\Phi}_4={\bar\phi}_5=\frAc{b}{1-\oMEga}, \lAbel{EQ:fOurTEenb}\\ {\bAr\phi}_{N+\tILdE{n}}={\bar\phi}_n, \label{eq:fourTeEnc} \

}{q})2\pi$ ($r,~q$, co-pri me integer s, $q >0$ ),in whi ch c ase both the a m plit ude and phase are repe atedat inte r va ls of $\tild e {N } = q N $la tti ce si tes.Onthe oth er hand, i f $ \f rac{\delta}{ 2 \p i}$ is irr ati onal, (\[eq: ten \]) re pr ese n ts an [* incom mensur a te*]{} standing w a ve sin c e the p h a se of$\psi_n$ never re p ea t s itself, butcomesar b it r a ril y c lose to an ypreas s igned v a lu e $ \ba r {\delta}$ for some $n$ o r ot her..5 cm In sectio n (5) ,w e c onsider for the sake ofconcre t eness t w o parti cular[*c omm ensu r at e* ]{} s o lut i on s w i th$\delta= \p i$ andwith $ N = 2,~3 $ r espe ctive ly. The solut ion for $N= 2,~\d elta= \pi~ (\ tilde {N}=4) $ rea ds (up to transla tion by a lat tic esit e) $$\ b egin{a lig ned } {\bar \phi}_0 = -{\ ba r \ p hi }_2=b\label{eq:thi rt e e na },\\ {\b ar\phi } _1 ={ \ bar\phi} _3 =0\lab e l {eq:t hirt e en b},\\ {\ bar\ph i }_ {n +\tilde {N }}={\b ar \ph i}_ n, \l a bel{ eq:thi rteenc}\ end{a l igned}$$ wher e $$b=\frac{\g a mm a } {\ g amma +\o mega}, \lab el{e q :thi rtee n d} $$a nd wh ere $ \g a mm a $ is to satisfy the c onsist encycondition $$\ gamma>\gam m a _ 1(\lambd a)|_ { N= 2 ,~\delta=\pi}=-\om ega. \labe l {eq:thir teene }$$ .5c m Recallt h at in th ispap erwer e st rict to $\ome g a <-2$ ,while s imi lar res ult s a reobt ai ned for $ \omega>2 $. I nasim ilarm anner, o ne ha s, fo r $N= 3 $, $\d elta= \pi$ ( pe r iod icity $ \ ti l d e{N} =6 $) $$ \be gi n{ali gned } {\ bar\phi }_0=-{\ba r\p h i}_3 =b ,\label{ eq:fourteena} \\ {\bar\phi }_ 1=- {\bar\ p h i}_2=-{\ bar\phi}_4={\bar\phi}_5 = \frac{b }{1 -\ome ga}, \label{e q:f ourtee nb} \ \ {\ba r\phi} _{n+\ ti lde { N }}={\ b a r\ phi }_ n, \label{ e q :fo urtee nc } \

}{q})2\pi$ ($r,~q$,_co-prime integers,_$q>0$), in which case_both the_amplitude_and phase_are_repeated at intervals_of $\tilde{N}=q N$_lattice sites. On the_other hand, if_$\frac{\delta}{2\pi}$_is irrational, (\[eq:ten\]) represents an [*incommensurate*]{} standing wave since the phase of $\psi_n$ never_repeats_itself, but_comes_arbitrarily_close to any preassigned value_$\bar{\delta}$ for some $n$ or_other..5cm In_section (5), we consider for the sake of_concreteness_two particular [*commensurate*]{}_solutions with $\delta=\pi$ and with $N=2,~3$ respectively. The solution_for $N=2,~\delta=\pi~(\tilde{N}=4)$ reads (up to translation_by a lattice_site) $$\begin{aligned} {\bar\phi}_0=-{\bar\phi}_2=b\label{eq:thirteena},\\ {\bar\phi}_1={\bar\phi}_3=0_\label{eq:thirteenb},\\ {\bar\phi}_{n+\tilde{N}}={\bar\phi}_n,_\label{eq:thirteenc}\end{aligned}$$ where $$b=\frac{\gamma}{\gamma+\omega}, \label{eq:thirteend}$$ and_where $\gamma$ is to satisfy the_consistency condition $$\gamma>\gamma_1(\lambda)|_{N=2,~\delta=\pi} =-\omega. \label{eq:thirteene}$$ .5cm Recall_that in this paper we restrict to_$\omega<-2$, while similar results are obtained_for $\omega>2$. In a similar_manner, one_has, for $N=3$, $\delta=\pi$ (periodicity_$\tilde{N}=6$) $$\begin{aligned} {\bar\phi}_0=-{\bar\phi}_3=b, \label{eq:fourteena}\\ {\bar\phi}_1=-{\bar\phi}_2=-{\bar\phi}_4={\bar\phi}_5=\frac{b}{1-\omega}, \label{eq:fourteenb}\\ {\bar\phi}_{n+\tilde{N}}={\bar\phi}_n,_\label{eq:fourteenc} \

$E(B-V) = 0.003 \pm 0.002$ [@Crawford.75; @Taylor.80] and an average distance modulus of $(m-M)_{0} = 3.33 \pm 0.01$ [@Perryman.98]. Comparison of Period-Age Data with Models ----------------------------------------- In figure \[fig:BVPeriod\] we compare the $(B-V)_{0}$ vs. Period relation for M37 to each of the four clusters discussed above. In figure \[fig:VIPeriod\] we show the $(V-I_{C})_{0}$ vs. Period relation. The convergence of stellar rotation periods into a sequence for $(B-V)_{0} < 1.3$ is clearly seen in M37 and the Hyades. A sequence may also be present in the Pleiades and M34 (this is more clearly seen in the less complete $BV$ data for M34). The data for NGC 2516 is incomplete for stars hotter than $(B-V)_{0} \sim 1.3$ as the I07 survey was focused on very low mass stars. The formation of such a sequence can be explained by an angular momentum loss law that is a steep function of the angular frequency. Typically modelers adopt a modified @Kawaler.88 $N = 1.5$ wind law [@Chaboyer.95], where $N$ is a parameter used in modeling the geometry of the coronal magnetic field of the star and can vary between 0 and 2, with $N = 3/7$ for a dipolar field and $N = 2$ for a purely radial field. The general law has the form: $$\frac{dJ}{dt} = \left\{ \begin{array}{ll} f_{k}K_{w}\omega^{1 + 4N/3}\left(\frac{R}{R_{\odot}}\right)^{2-N}\left(\frac{M}{M_{\odot}}\right)^{-N/3}\left(\frac{\dot{M}}{10^{-14}M_{\odot}{\rm yr}^{-1}}\right)^{1 - 2N/3}, & \omega < \omega_{sat} \\ f_{k}K_{w}\omega\omega_{crit}^{4N/3}\left(\frac{R}{R_{\odot}}\right)^{2-N}\left(\frac{M}{M_{\

$ E(B - V) = 0.003 \pm 0.002 $ [ @Crawford.75; @Taylor.80 ] and an average distance modulus of $ (m - M)_{0 } = 3.33 \pm 0.01 $ [ @Perryman.98 ]. Comparison of Period - Age Data with Models ----------------------------------------- In figure   \[fig: BVPeriod\ ] we compare the $ (B - V)_{0}$ vs. Period relation back for M37 to each of the four bunch discussed above. In figure   \[fig: VIPeriod\ ] we picture the $ (vanadium - I_{C})_{0}$ vs. Period relation. The convergence of leading rotation periods into a sequence for $ (B - V)_{0 } < 1.3 $ is intelligibly seen in M37 and the Hyades. A sequence may besides be present in the Pleiades and M34 (this is more clearly seen in the less complete $ BV$ data for M34). The datum for NGC 2516 is incomplete for stars hotter than $ (B - V)_{0 } \sim 1.3 $ as the I07 sketch was focused on very low aggregate stars. The formation of such a sequence can be explain by an angular momentum loss law that is a exorbitant function of the angular frequency. Typically modeler adopt a modified @Kawaler.88 $ N = 1.5 $ wind law [ @Chaboyer.95 ], where $ N$ is a parameter used in model the geometry of the coronal magnetic field of the star and can vary between 0 and 2, with $ N = 3/7 $ for a dipolar field and $ N = 2 $ for a strictly radial battlefield. The general law has the form: $ $ \frac{dJ}{dt } = \left\ { \begin{array}{ll } f_{k}K_{w}\omega^{1 + 4N/3}\left(\frac{R}{R_{\odot}}\right)^{2 - N}\left(\frac{M}{M_{\odot}}\right)^{-N/3}\left(\frac{\dot{M}}{10^{-14}M_{\odot}{\rm yr}^{-1}}\right)^{1 - 2N/3 }, & \omega < \omega_{sat } \\ f_{k}K_{w}\omega\omega_{crit}^{4N/3}\left(\frac{R}{R_{\odot}}\right)^{2 - N}\left(\frac{M}{M_{\

$E(B-G) = 0.003 \pm 0.002$ [@Crawford.75; @Taylor.80] xnd an average bustancx modulhs of $(m-M)_{0} = 3.33 \pm 0.01$ [@Perryman.98]. Comparison of Pwriod-Qge Data with Models ----------------------------------------- In figure \[fih:BVPerioe\] we xompare thx $(B-V)_{0}$ vs. Izriod velatnoi for M37 to each of the fogr clusters divcjsded above. In figure \[fig:VIPeriod\] we srow the $(V-L_{C})_{0}$ vs. Period rglatipg. The bokvergence of stellar rotation pediods iito a sequence gor $(B-V)_{0} < 1.3$ is clearly seen ij M37 wnd the Hyades. A sfquence may alsj be present kn the Plepcdes and M34 (jhis is more clearly seen in the less complete $VV$ dahd for M34). The data for NGC 2516 is incomplate for stars hotter bhan $(U-V)_{0} \sum 1.3$ as the I07 survey wes focused on very ljw mass sdaxs. The formation of suxh a sexuenwe cxb bd espkajned bj ai angular mkmentum losw law that is a stetp slmction of ths angujaw frequency. Typically modelers adopt a kodjfied @Kawaler.88 $N = 1.5$ wind oaw [@Chaboyer.95], where $N$ ls a paraieter used in modeling the geometry of the coronan magiegic fleld if the star and can vary between 0 and 2, with $N = 3/7$ rot s dipolar fielb and $N = 2$ for a lugeki radial field. The gzhedal law has the fogm: $$\frac{qJ}{dt} = \left\{ \begyn{artay}{ll} f_{k}K_{w}\omega^{1 + 4N/3}\left(\frqc{R}{R_{\odot}}\righn)^{2-N}\ledt(\frac{M}{M_{\odot}}\right)^{-N/3}\peft(\frac{\dot{O}}{10^{-14}M_{\odpt}{\rm ur}^{-1}}\right)^{1 - 2N/3}, & \omega < \omegc_{sat} \\ f_{k}K_{w}\omegw\omega_{crif}^{4V/3}\left(\frac{R}{R_{\odot}}\rkghn)^{2-N}\lext(\frac{M}{M_{\

$E(B-V) = 0.003 \pm 0.002$ [@Crawford.75; @Taylor.80] average modulus of = 3.33 \pm Data Models ----------------------------------------- In \[fig:BVPeriod\] we compare $(B-V)_{0}$ vs. Period relation for M37 each of the four clusters discussed above. In figure \[fig:VIPeriod\] we show the vs. Period relation. The convergence of stellar rotation periods into a sequence for < is seen M37 and the Hyades. A sequence may also be present in the Pleiades and M34 (this more clearly seen in the less complete $BV$ for M34). The data NGC 2516 is incomplete for hotter $(B-V)_{0} \sim as I07 was focused on low mass stars. The formation of such a sequence can be explained by an angular momentum loss that is function of angular Typically adopt a modified = 1.5$ wind law [@Chaboyer.95], where parameter used in modeling the geometry of the magnetic field the star and can vary between and 2, with $N = 3/7$ for a field and $N = 2$ for a purely radial field. The general law has the = \left\{ \begin{array}{ll} f_{k}K_{w}\omega^{1 4N/3}\left(\frac{R}{R_{\odot}}\right)^{2-N}\left(\frac{M}{M_{\odot}}\right)^{-N/3}\left(\frac{\dot{M}}{10^{-14}M_{\odot}{\rm yr}^{-1}}\right)^{1 - & < \\

$E(B-V) = 0.003 \pm 0.002$ [@Crawford.75; @Taylor.80] and an aVerage distAnce mOduLus Of $(M-M)_{0} = 3.33 \pm 0.01$ [@perrYman.98]. Comparison OF PerIod-Age Data with Models ----------------------------------------- In FigurE \[fIG:BVPERiOd\] we cOmpare tHE $(B-v)_{0}$ VS. PeRiOd RelAtIOn For M37 tO eaCh of the Four clusteRs dIsCussed above. IN FiGure \[fig:VIPEriOd\] we show the $(V-i_{C})_{0}$ vS. PerioD rElaTIon. ThE coNvergEnce of STellar Rotation pErIOds intO A sequenCE FoR $(B-V)_{0} < 1.3$ iS clearly seen in M37 anD ThE hyades. A sequencE may alSo BE pRESenT in The PleiadeS aNd M34 (thIS is more CLeARLY seEN in the less comPlete $BV$ data FOr M34). the datA fOr Ngc 2516 is incOmpleTe FOr sTars hotter tHan $(B-v)_{0} \sim 1.3$ as the i07 surveY Was focuSEd on verY low maSs sTarS. The FOrMaTioN oF SucH A sEquENce Can be expLaInEd by aN angULAR MomeNtuM losS law tHat is a steep fuNctIon oF The AngulAr freQuenCy. typicAlly moDelerS aDopt a modified @KaWaleR.88 $N = 1.5$ wind law [@chaBoYer.95], WhEre $N$ iS A paramEteR usEd in modEling thE GeoMeTRY Of The coronal magnetic FiELD oF the star And can VArY bETween 0 and 2, WiTh $N = 3/7$ For a DIPolar FielD AnD $N = 2$ for a puRely raDIaL fIeld. The GeNeral lAw Has The Form: $$\fRAc{dJ}{Dt} = \left\{ \Begin{arrAy}{ll} f_{K}k_{w}\omega^{1 + 4N/3}\left(\frAC{R}{R_{\odot}}\right)^{2-N}\LEfT(\FRaC{m}{M_{\odOt}}\rIght)^{-N/3}\left(\frAc{\doT{m}}{10^{-14}M_{\odOt}{\rm YR}^{-1}}\rIghT)^{1 - 2n/3}, & \omegA < \omegA_{sAT} \\ f_{K}k_{w}\omega\omega_{crit}^{4N/3}\leFt(\Frac{R}{R_{\Odot}}\rIght)^{2-N}\left(\frac{m}{M_{\

$E(B-V) = 0.003 \pm 0.002 $ [@Crawfo rd.75 ; @ Tay lo r.80 ] an d an average d i stan ce modulus of $(m-M)_{ 0} =3. 3 3 \p m 0 .01$[@Perry m an . 9 8]. Co mpa ri s on of P eri od-AgeData withMod el s ---------- - -- ---------- --- ------------ --- In f ig ure \[fig :BV Perio d\] we compar e the $(B -V ) _{0}$v s. Peri o d r elat ion for M37 to ea c ho f the four clu stersdi s cu s s edabo ve. In fig ur e \[f i g:VIPer i od \ ] wes how the $(V-I _{C})_{0}$v s.Period r ela t ion. The c on v erg ence of ste llar rotation perio d s intoa sequen ce for $( B-V )_{0 } < 1 .3$ i s cl e ar lys een in M37an dthe H yade s . A seq uen ce m ay al so be present in the Ple iades andM34(t his i s more clea rl y seen in the l esscomplete$BV $dat afor M 3 4). Th e d ata for NG C 2516i s i nc o m p le te for stars hotte rt h an $(B-V)_ {0} \s i m1. 3 $ as the I 07surv e y wasfocu s ed on very low m a ss s tars. Th e form at ion of such a se quence can beexpla i ned by an angu l ar momentum l o ss l aw that is a steep fu ncti o n of the an gul a r fre quenc y. Ty p ically modelers ado pt a mod ified @Kawaler.88$N = 1.5$w i n d law [@ Chab o ye r .95], where $N $ isa paramete r used in mode ling the geometry o f the co ron almag net i c f ield of the s t a r an dcan var y b etween0 a nd2,wit h$N = 3/7$ for a d ip ol ar f iel d and $N = 2$fo r a p ure ly ra d ial fi eld.Thege ne r allaw has th e form :$$ \fra c{d J} {dt}= \l e ft\ { \begi n{array}{ ll} f_ {k }K_{w}\ omega^{1 + 4N /3 }\left(\fr ac {R} {R_{\o d o t}}\righ t)^{2-N}\left(\frac{M}{ M _{\odot }}\ right )^{- N/3}\left (\f rac{\d ot{ M }}{10^ {-14}M _{\od ot }{\ r m yr}^ { - 1} }\r ig ht)^{1 - 2 N / 3}, & \o me ga < \omega _{sat} \\ f_{k } K_{ w}\omega\omeg a_{ crit } ^ {4 N/3 } \l e ft( \f r ac{ R } {R_{\odot}}\rig ht)^{2-N}\ le f t( \frac{M}{M _ {\

$E(B-V)_= 0.003_\pm 0.002$ [@Crawford.75; @Taylor.80]_and an_average_distance modulus_of_$(m-M)_{0} = 3.33_\pm 0.01$ [@Perryman.98]. Comparison_of Period-Age Data with_Models ----------------------------------------- In figure \[fig:BVPeriod\] we_compare_the $(B-V)_{0}$ vs. Period relation for M37 to each of the four clusters discussed_above._In figure \[fig:VIPeriod\]_we_show_the $(V-I_{C})_{0}$ vs. Period relation. The_convergence of stellar rotation periods_into a_sequence for $(B-V)_{0} < 1.3$ is clearly seen_in_M37 and the_Hyades. A sequence may also be present in the_Pleiades and M34 (this is more_clearly seen in_the_less_complete $BV$ data for_M34). The data for NGC 2516_is incomplete for stars hotter than_$(B-V)_{0} \sim 1.3$ as the I07 survey_was focused on very low mass_stars. The formation of such a_sequence can_be explained by an angular_momentum loss law_that is_a steep function_of the angular frequency. Typically modelers_adopt a modified_@Kawaler.88 $N = 1.5$ wind law_[@Chaboyer.95],_where $N$ is_a_parameter_used in_modeling the geometry_of_the coronal_magnetic_field of the star and can_vary_between 0 and 2, with $N =_3/7$ for a dipolar_field_and $N = 2$_for a purely radial field._The general law has the form:_$$\frac{dJ}{dt} =_\left\{ \begin{array}{ll} _ f_{k}K_{w}\omega^{1 + 4N/3}\left(\frac{R}{R_{\odot}}\right)^{2-N}\left(\frac{M}{M_{\odot}}\right)^{-N/3}\left(\frac{\dot{M}}{10^{-14}M_{\odot}{\rm yr}^{-1}}\right)^{1 - 2N/3}, & \omega_< \omega_{sat} \\ _f_{k}K_{w}\omega\omega_{crit}^{4N/3}\left(\frac{R}{R_{\odot}}\right)^{2-N}\left(\frac{M}{M_{\

, in particular, global and domestic systemically important banks (G-SIBs or D-SIBs). The BCBS recommends increased capital requirements for SIFIs – the so called “SIFI surcharges” [@BIS:2010aa; @Georg:2011aa]. They propose that SR should be measured in terms of the impact that a bank’s failure can have on the global financial system and the wider economy, rather than just the risk that a failure could occur. Therefore they understand SR as a global, system-wide, loss-given-default (LGD) concept, as opposed to a probability of default (PD) concept. Instead of using quantitative models to estimate SR, Basel III proposes an indicator-based approach that includes the size of banks, their interconnectedness, and other quantitative and qualitative aspects of systemic importance. There is not much literature on the problem of dynamically re-shaping network topology so that networks adapt over time to function optimally in terms of stability and efficiency. A major problem in NW-based SR management is to provide agents with incentives to re-arrange their local contracts so that global (system-wide) SR is reduced. Recently, it has been noted empirically that individual transactions in the interbank market alter the SR in the total financial system in a measurable way [@Poledna:2014aa; @Poledna:2015aa]. This allows an estimation of the marginal SR associated with financial transactions, a fact that has been used to propose a tax on systemically relevant transactions [@Poledna:2014aa]. It was demonstrated with an agent-based model (ABM) that such a tax – the [*systemic risk tax*]{} (SRT) – is leading to a dynamical re-structuring of financial networks, so that overall SR is substantially reduced [@Poledna:2014aa]. In this paper we study and compare the consequences of two different options for the regulation of SR with an ABM. As an example for the first option we study Basel III with capital surcharges for G-SIBs and compare it with an example for the second option – the SRT that leads to a self-organized re-structuring of financial networks. A number of ABMs have been used recently to study interactions between the financial system and the real economy, focusing on destabilizing feedback loops between the two sectors [@Delli-Gatti:2009aa; @Battiston:2012ab; @Tedes

, in particular, global and domestic systemically important banks (G - sibling or vitamin d - SIBs). The BCBS recommends increased capital necessity for SIFIs – the so called “ SIFI surcharge ” [ @BIS:2010aa; @Georg:2011aa ]. They propose that SR should be measured in terms of the impingement that a bank ’s failure can get on the global financial system and the wider economy, rather than just the hazard that a failure could occur. Therefore they sympathize SR as a global, system - wide, loss - given - nonpayment (LGD) concept, as opposed to a probability of default (PD) concept. Instead of using quantitative models to calculate SR, Basel III proposes an indicator - based approach that includes the size of banks, their interconnection, and early quantitative and qualitative expression of systemic importance. There is not much literature on the problem of dynamically re - shaping network topology so that networks adjust over time to serve optimally in terms of stability and efficiency. A major problem in NW - free-base SR management is to provide agents with incentives to re - dress their local contracts so that global (system - wide) SR is reduced. Recently, it has been note empirically that individual transactions in the interbank market alter the SR in the total financial system in a measurable means [ @Poledna:2014aa; @Poledna:2015aa ]. This allows an estimation of the marginal SR associated with fiscal transactions, a fact that has been use to nominate a tax on systemically relevant transactions [ @Poledna:2014aa ]. It was demonstrated with an agent - based exemplar (ABM) that such a tax – the [ * systemic risk tax * ] { } (SRT) – is leading to a dynamical rhenium - structuring of financial networks, so that overall SR is substantially reduced [ @Poledna:2014aa ]. In this newspaper we study and compare the consequence of two different options for the rule of SR with an ABM. As an example for the first option we study Basel III with das kapital surcharges for G - SIBs and compare it with an case for the second option – the SRT that leads to a self - organize re - structuring of fiscal networks. A number of ABMs have been used recently to learn interactions between the financial system and the real economy, focusing on destabilizing feedback cringle between the two sectors [ @Delli - Gatti:2009aa; @Battiston:2012ab; @Tedes

, in particular, global and dumestic systemieqlly ikportaht banks (G-SIBs or D-SIBs). The BCBS reclmnends increased capital reqjirements for SIFUs – uhe so called “SIFM surchavyes” [@BJD:2010aa; @Yeirg:2011aa]. They prokose that SR should be meavufeb in terms of the impact that a bank’f failute can have on tre gkjbal financial system and the wider ecknomy, rether than just the risk that a failure cluld occur. Therefore tjey understqnd FE as a globau, system-wide, loss-given-sefault (LGD) concept, as opposed go a 'robability od dfxault (PD) coicept. Pnstead of uslmg quattitatife models to extikatw SR, Basel III proposxs an indicator-based approach tkat includes the size od banns, tveir untdrckniecfednesd, aid other quzntitative qnd qualitative asptctf of systemic jmportwnse. There is not much literature on the pgoblsm of dynamically re-shaping network topology so that getworks adapt over time to function optimally in terma of sbabiuutj and efficiency. A major problem in NW-based SW msnsgement is to irovide agents witn lnvgntives to re-atrange cgejr local contracts so thaj globql (system-ride) SR is reduced. Recently, it yas been notvd enpirically that inbividual trausactipns im the interbank market clter fhe SR in tje total rknancial system kn s keasurable way [@Poledna:2014aa; @Pjledna:2015aa]. Vhis cllows av esjimatiog of the mwrginal SR associated wihh fiuancidl transachions, a fact that has been used to propose a jax on systemieally velevant transastions [@Poledna:2014ca]. It waf demunstrated sith an agent-based model (ABM) thdj such a tax – the [*syseemix riwk tax*]{} (RFT) – is leading to a dynciicao re-structuring of fivzncial networks, so that overall ST ir stbdtentiajny reduced [@Pmledva:2014ax]. On thks paper we rtudu and compare the cotseqhences of two diffrrcnt optiobs for tre regulation of SR with an ABM. As ai exam'le fot tre first option we study Basel III with caiital surchardes nor D-SIBs and eompare it with an example for the seconv option – the SRT that oeads to a self-orgauieed re-structucing os financidl networks. A number of ABMs have beek used recently to stusy intaractlons between the financial system and the real economy, focusing on destabulizinj seedback lokps netwevn che two sqctocs [@Delli-Gatti:2009aa; @Bantiston:2012ab; @Tedes

, in particular, global and domestic systemically (G-SIBs D-SIBs). The recommends increased capital so “SIFI surcharges” [@BIS:2010aa; They propose that should be measured in terms of impact that a bank’s failure can have on the global financial system and wider economy, rather than just the risk that a failure could occur. Therefore understand as global, loss-given-default (LGD) concept, as opposed to a probability of default (PD) concept. Instead of using quantitative to estimate SR, Basel III proposes an indicator-based that includes the size banks, their interconnectedness, and other and aspects of importance. is much literature on problem of dynamically re-shaping network topology so that networks adapt over time to function optimally in terms stability and major problem NW-based management to provide agents to re-arrange their local contracts so SR is reduced. Recently, it has been noted that individual in the interbank market alter the in the total financial system in a measurable [@Poledna:2014aa; @Poledna:2015aa]. This allows an estimation of the marginal SR associated with financial transactions, a has been used to a tax on relevant [@Poledna:2014aa]. was with an model (ABM) that such a tax – the [*systemic risk tax*]{} – is leading to a dynamical re-structuring of financial networks, overall is substantially reduced In this paper we and the consequences of two for regulation an As example for the first we study Basel III with surcharges for G-SIBs and for the second option – the SRT that to a self-organized re-structuring of financial networks. number of ABMs have been used recently to study interactions between the system and economy, focusing on destabilizing feedback loops between the sectors [@Delli-Gatti:2009aa; @Battiston:2012ab; @Tedes

, in particular, global and domeStic systemIcallY imPorTaNt baNks (G-sIBs or D-SIBs). The bcBS rEcommends increased capiTal reQuIRemeNTs For SIfIs – the sO CaLLEd “SiFi sUrcHaRGeS” [@BIS:2010aA; @GeOrg:2011aa]. ThEy propose tHat sR Should be measUReD in terms of The Impact that a bAnk’S failuRe Can HAve on The GlobaL finanCIal sysTem and the WiDEr econOMy, ratheR THaN jusT the risk that a failURe COuld occur. ThereFore thEy UNdERStaNd Sr as a global, SyStem-wIDe, loss-gIVeN-DEFauLT (LGD) concept, as Opposed to a pRObaBility Of DefAUlt (PD) cOncepT. INSteAd of using quAntiTative modEls to eSTimate Sr, basel IIi propoSes An iNdicAToR-bAseD aPProACh ThaT IncLudes the SiZe Of banKs, thEIR INterConNectEdnesS, and other quanTitAtivE And QualiTativE aspEcTs of sYstemiC impoRtAnce. There is not mUch lIterature On tHe ProBlEm of dYNamicaLly Re-sHaping nEtwork tOPolOgY SO ThAt networks adapt oveR tIME tO functioN optimALlY iN Terms of sTaBilIty aND EfficIencY. a mAjor probLem in Nw-BaSeD SR manaGeMent is To ProVidE agenTS witH incenTives to rE-arraNGe their local coNTracts so that gLObAL (SySTem-wIde) sR is reduced. receNTly, iT has BEeN noTEd empIricaLlY ThAT individual transactIoNs in thE inteRbank market alTer the SR in THE Total finAnciAL sYStem in a measuraBle waY [@Poledna:2014aa; @pOledna:2015aa]. this aLlows an eStimation OF The margiNal sR aSsoCiaTED wIth financial tRANsacTiOns, a facT thAt has beEn uSed To pRopOsE a tax on syStemicalLy ReLeVaNt tRansaCTions [@PolEdNa:2014aA]. IT waS demoNStrateD with An agEnT-bASed Model (ABm) ThAT Such A tAx – The [*sYstEmIc risK tax*]{} (srT) – iS leadinG to a dynamIcaL Re-stRuCtUring of Financial netwOrKs, so that ovErAll sR is suBSTantiallY reduced [@Poledna:2014aa]. In this PAper we sTudY and cOmpaRe the consEquEnces oF twO DifferEnt optIons fOr The REGulatION oF SR WiTh an ABM. As aN EXamPle foR tHe fiRst optiOn we study Basel III wITh cApital surcharGes For G-sibs And COmPAre It WIth AN Example for the seCond option – ThE sRt that leads TO a sElF-organiZed re-stRuctuRIng of fiNancial neTworks. A nuMbEr of abms hAve been useD recentlY to study iNTeracTIoNs betWeeN the fiNaNciAl sysTem and THe rEal ecOnomy, fOcUsing oN destAbIlizing fEedback loops between the tWo sectOrs [@DeLli-gatti:2009aa; @BaTtiSTon:2012Ab; @Tedes

, in particular, global an d domestic syst emi cal ly imp orta nt banks (G-SI B s or D-SIBs). The BCBS rec ommen ds incr e as ed ca pital r e qu i r eme nt sfor S I FI s – t heso call ed “SIFI s urc ha rges” [@BIS: 2 01 0aa; @Geor g:2 011aa]. They pr oposeth atS R sho uld be m easure d in te rms of th ei mpactt hat a b a n k’ s fa ilure can have on th e global financ ial sy st e ma n d t hewider econ om y, ra t her tha n j u s t th e risk that afailure cou l d o ccur.Th ere f ore th ey un de r sta nd SR as aglob al, syste m-wide , loss-g i ven-def ault ( LGD ) c once p t, a s o pp o sed to ap rob abilityof d efaul t (P D ) c once pt. Ins teadof using quan tit ativ e mo delsto es tima te SR,BaselIII p ro poses an indica tor- based app roa ch th at incl u des th e s ize of ban ks, the i r i nt e r c on nectedness, and ot he r qu antitati ve and qu al i tative a sp ect s of s ystem ic i m po rtance. There is n ot much l iterat ur e o n t he pr o blem of dy namicall y re- s haping network topology so t h at n et w orks ad apt over ti me t o fun ctio n o pti m allyin te rm s o f stability and effi ci ency.A maj or problem in NW-basedS R manageme nt i s t o provide agent s wit h incentiv e s to re- arran ge their local co n t racts so th atglo bal ( sy stem-wide) SR i s re du ced. Re cen tly, it ha s b een no te d empiric ally tha tin di vi dua l tra n sactions i n t he in terba n k mark et al terth eS R i n the t o ta l fina nc ia l sy ste min ameas u rab le way[@Poledna :20 1 4aa; @ Po ledna:2 015aa]. Thisal lows an es ti mat ion of t he margi nal SR associated withf inancia l t ransa ctio ns, a fac t t hat ha s b e en use d to p ropos ea t a x on s y s te mic al ly relevan t tra nsact io ns [ @Poledn a:2014aa]. It wasd emo nstrated with an age n t -b ase d m o del ( A BM) t hat such a tax– the [*sy st e mi c risk tax * ]{} ( SRT) –is lead ing t o a dyna mical re- structuri ng off i nan cial netwo rks, sothat over a ll SR is subs tan tially r edu ced [ @Poled n a:2 014aa ]. In t his pa per w estudy an d compare the consequen ces of twodif ferent op tio n s f or the re gula tion of SR wi than AB M.A s anexam p le fo r thefirs t option w e s tud y Ba sel III wit h c api tal s urc h argesforG-SIBs and compar e it with an ex ampl e for th e sec on d option – the SR Tt h at leads t o a self-or ganizedre - struc turing of fi nancial n et w orks.A nu mbe r of ABMs ha ve been us ed r e cently tost udy in teract i onsb e tween the financ ial s y s tem a n d t he re al econom y , fo cusing ondestabilizi ng fee dbac k loo ps betw ee n thetwo s ectors [@D e lli-Gatti :2009 aa; @Ba tt isto n:2 012ab; @Te d e s

, in_particular, global_and domestic systemically important_banks (G-SIBs_or_D-SIBs). The_BCBS_recommends increased capital_requirements for SIFIs_– the so called_“SIFI surcharges” [@BIS:2010aa;_@Georg:2011aa]._They propose that SR should be measured in terms of the impact that a_bank’s_failure can_have_on_the global financial system and_the wider economy, rather than_just the_risk that a failure could occur. Therefore they_understand_SR as a_global, system-wide, loss-given-default (LGD) concept, as opposed to a_probability of default (PD) concept. Instead_of using quantitative_models_to_estimate SR, Basel III_proposes an indicator-based approach that includes_the size of banks, their interconnectedness,_and other quantitative and qualitative aspects of_systemic importance. There is not much literature_on the problem of dynamically_re-shaping network_topology so that networks adapt_over time to_function optimally_in terms of_stability and efficiency. A major problem_in NW-based SR_management is to provide agents with_incentives_to re-arrange their_local_contracts_so that_global (system-wide) SR_is_reduced. Recently,_it_has been noted empirically that individual_transactions_in the interbank market alter the SR_in the total financial_system_in a measurable way_[@Poledna:2014aa; @Poledna:2015aa]. This allows an_estimation of the marginal SR associated_with financial_transactions, a_fact that has been used to propose a tax on systemically_relevant transactions [@Poledna:2014aa]. It was demonstrated_with an agent-based model_(ABM) that_such_a tax –_the_[*systemic risk_tax*]{} (SRT) – is leading to a_dynamical re-structuring_of financial networks, so that overall_SR is substantially reduced_[@Poledna:2014aa]. In_this paper we study and compare_the consequences of two different options_for the regulation of SR_with_an_ABM. As an example for_the first option we study Basel_III with capital_surcharges for G-SIBs and compare it with_an_example for the second option –_the_SRT that leads to a self-organized_re-structuring_of_financial networks. A number of_ABMs have been used recently to_study interactions between the financial system and the real_economy, focusing on_destabilizing feedback loops between the_two_sectors_[@Delli-Gatti:2009aa; @Battiston:2012ab; @Tedes